TECH

Why the Pentagon pushed AI to the limit during the war with Iran?

The Pentagon has used artificial intelligence in the war with Iran on a scale that has virtually stretched the classified computing capacity available in the United States to the limit. The Maven system, developed by Palantir, gathered information from satellites, sensors and other intelligence sources to support target identification and speed up military decisions. The operation represented a breakthrough in integrating AI into the U.S. military. Technologies that, a few years ago, still presented difficulties in reliably identifying military objectives have become integrated into the planning and execution of real operations. The experience is now serving as a reference for the adoption of similar systems by allied countries. According to an analysis published by Reuters on October 30, Cameron Stanley, head of digital and artificial intelligence at the Pentagon, confirmed that the operation against Iran marked the first full use of Maven's analysis and target identification capabilities. Demand for classified computing resources quickly reached the limit of available capacity. reuters.com Maven made AI part of military operations...Maven combines information from different surveillance and reconnaissance systems. The platform allows you to analyze images, cross-reference sensor data and identify potential targets, tasks that previously required a longer intervention from military analysts. Palantir's technology integrates into an ecosystem that includes artificial intelligence models from companies such as Anthropic, OpenAI, Google and Microsoft. SpaceX also supplies technology used in U.S. military operations, although the roles and levels of integration of these suppliers are not necessarily equivalent. The change is significant because AI is no longer limited to experiences or administrative tasks. He became directly involved in the processing of the information that underpins operational decisions. This evolution does not mean, however, that systems can autonomously decide all attacks. Palantir says decisions about the use of lethal force remain human responsibility. In March, the Pentagon decided to institutionalize Maven as one of the main artificial intelligence systems of the Armed Forces. The decision sought to secure funding and accelerate the adoption of the platform by the different military structures. Analysis errors expose military AI risks..The speed of processing does not eliminate the problems associated with the quality of the data or the interpretation of the results. One of the most relevant episodes occurred when an incorrect analysis, produced using artificial intelligence, led the US military to consider the approach of a Chinese merchant ship. The vessel was suspected of carrying components related to nuclear weapons, but the information turned out to be wrong. According to the report cited by Reuters, the mistake was identified before US forces advanced with the intervention. The episode demonstrates how an incorrect conclusion can trigger operational decisions with diplomatic and military consequences. It is not demonstrated that the analysis was produced specifically by Maven. Another case that raised questions was the bombing of a school in Iran during the first day of the offensive. The US authorities attributed the incident to the use of outdated maps and information, without identifying the analysis produced by AI as the cause of the error. The two episodes illustrate distinct problems. A model can misinterpret available information, but it can also produce technically coherent results from data that is already wrong. In military operations, the combination of these factors increases the importance of human verification, especially when an assessment may lead to the use of weaponry. Technological dependence worries allies of the United States...The military's use of artificial intelligence is also creating new dependencies between governments and tech companies. The United States is seeking to help European, Asian and Middle Eastern allies adopt tools similar to those used during the war with Iran. However, some countries fear losing autonomy if access to these platforms depends on decisions made in Washington or by private providers. The issue became particularly relevant after the conflict between the Pentagon and Anthropic over the limits imposed on the military use of their models. TecheNet had already addressed this dispute in an article on the Pentagon'S new AI agreements that excluded Anthropic, published in May 2026. Palantir sought to address similar concerns about Maven. Louis Mosley, the company's head of UK and Europe, told Reuters that customers who have purchased the platform retain operational control and that there is no mechanism that would allow the company or the US government to remotely disable the system. NATO acquired Maven in 2025 and declared that the platform achieved full technical operational capability in June 2026. Countries such as Germany and France, however, have expressed reservations about national adoption of the technology, not least for reasons of security and sovereignty. reuters.com More analysis capacity does not guarantee better results...The US experience shows that artificial intelligence can speed up the analysis of military information, but it does not necessarily solve the strategic problems of a conflict. Faster identification of potential targets can increase the pace of operations and the consumption of weaponry. It can also reduce the time available to verify information if human oversight procedures do not keep up with the speed of systems. The challenge goes beyond the technical capacity of the models. The Armed Forces will have to define how to validate recommendations, identify errors and assign responsibility when a decision supported by artificial intelligence causes unforeseen consequences. The use of Maven during the war with Iran thus represents a shift in the operational scale of military AI. Technology has come to occupy a central place in information processing and decision support, but experience has also exposed limitations that do not disappear with increasing computational capacity.

The Pentagon pushed AI to its limits during the conflict with Iran to drastically accelerate targeting decisions and scale up remote strike operations without deploying ground troops.

Key drivers behind the AI push: • Machine-speed target generation: The U.S. military used Palantir's Maven Smart System integrated with generative AI models (like Anthropic’s Claude) to aggregate massive streams of classified surveillance data. This reduced the time needed to process intelligence and build target lists from tens of thousands of human-hours down to seconds. • Massive strike scale: In the first 24 hours of the campaign starting February 28, 2026, the U.S. and Israel struck roughly 1,000 targets—double the initial shock-and-awe scale of the 2003 Iraq War. • Avoiding ground troops: Operating under strategic directives to minimize American boots on the ground, the administration relied on remote, AI-assisted air and missile power to project force rapidly. • The "AI-First" doctrine: Under the "AI Acceleration Strategy" and leadership pushing to eliminate internal technological constraints or "ideological guardrails," the Pentagon sought unconstrained use of commercial tech to maintain a perceived dominance over global adversaries like China. Controversies and limits: • Clashes with tech providers: The push triggered acute friction with tech companies like Anthropic, which attempted to restrict its models from powering autonomous lethal weapons or mass domestic surveillance, leading to temporary government supply-chain risk designations and pivots toward alternative defense contractors. • Battlefield errors and oversight: Rapid deployment raised intense legal and ethical debates regarding accountability in the "kill chain," highlighted by reported targeting errors—including strikes impacting civilian structures like an Iranian elementary school

mundophone

 

DIGITAL LIFE

Invisible SMS can drop calls on Android phones

A vulnerability in Android allows phone calls to be interrupted via an "invisible" SMS—a specially crafted text message that requires no user interaction, such as opening the message or receiving a notification. The flaw, dubbed "SMS of Death," affects how WAP Push messages are processed and can terminate ongoing calls, including those made to emergency services.

Security researcher Mark Blaszczyk disclosed the technical details of the issue following a communication process with Google that began in May 2026. The vulnerability was reproduced on a Google Pixel 9 and several Samsung Galaxy devices. Although Google has developed a fix, questions remain regarding its distribution across various manufacturers.

An invisible SMS crashes Android phones, cuts off calls, including to emergency services, and can trap users in a reboot loop, a researcher warns. The “SMS of Death” exploit code is now public, and the fix is underway.

Simply invisible SMS messages can make Android phones unusable, requiring a factory reset or an involved recovery procedure, a new security advisory warns.

“A single malformed WAP Push SMS crashes com.android.phone on Android, dropping active calls (including emergency calls), with zero interaction from the victim,” Mark Blaszczyk, a security researcher (@sourone), said in a report.

And attackers can already leverage SMSes to target victims using a proof-of-concept “SMS of Death” app, made available on GitHub – it remotely and silently sends a crafted 8-byte WAP Push PDU over SMS to a target handset.

The bug requires no user interaction, the exploit works remotely, and often bricks the device – recovery requires connecting to the phone using Android Debug Bridge (ADB), according to the researcher. The attack is completely invisible, and the SMS messages don’t appear in the inbox.

The author disclosed the bug responsibly in May, 2026. Google committed a fix on June 15 and pushed the patch to the Android Open Source Project (AOSP), but the fix has yet to reach all supported devices.

The critical vulnerability affects Android 12 through 16, meaning that many phones will never receive a fix.

GrapheneOS said the bug will be included in the upcoming November security bulletin. The critical flaw is patched in Android 14–17 with the October 2026 update. It has been assigned CVE-2026-58882.

How does the bug work? The entire payload is just eight bytes sent via SMS, and five of those bytes represent a single number – the header’s length.

When the phone receives the SMS and starts processing it automatically, it sees a valid value: 2,147,483,648, in a format known as long (0x80000000), which is used for large numbers.

However, Android then converts it to a smaller data type, a 32-bit integer (int), without checking the range, causing overflow. The number flips to a negative value (-2,147,483,648) after conversion, which makes no sense as a header length.

The negative value triggers a fatal error, killing com.android.phone, a component that handles calls.

WAP Push is a legacy system that lets mobile carriers send hidden instructions to phones without the user taking any action, such as prompting devices to download MMS or apply carrier settings.

The attack exploits a flaw in the processing of WAP Push messages—a mechanism used by mobile networks to transmit specific types of information to mobile phones.

According to security researcher Mark Blaszczyk, a specially crafted message can cause a failure in the `com.android.phone` process, which handles essential telephony functions on Android.

The issue stems from an incorrect interpretation of a message header's length. When the system receives certain malformed values, an exception can occur, causing the telephony process to crash.

A unique aspect of this vulnerability is that the attack requires no interaction from the victim. The message is processed automatically, without ever appearing in the user's inbox or triggering a visible notification.

In tests conducted by the researcher, the flaw caused a temporary disruption to telephony services—enough to terminate an active call. The issue could even affect calls to the European emergency number 112 if a malicious message is received while the call is in progress.

Repeated messages can prolong the disruption; the consequences are not necessarily limited to the interruption of a single call. An analysis published by eSecurity Planet outlines the risks associated with the repeated transmission of malicious messages, which can trigger further failures in telephone service.

Blaszczyk documented instances where repeated attacks rendered equipment temporarily unable to perform standard telephony functions. The researcher also noted that certain message-resending mechanisms can exacerbate the issue.

The extent of these consequences, however, depends on the specific hardware, the version of Android, and the conditions under which the attack occurs. The tests presented do not indicate that all Android phones exhibit the same behavior.

The vulnerability constitutes a denial-of-service attack. The findings provide no evidence that it allows for access to personal data, the installation of applications, or the complete takeover of a device.

The timeline of the vulnerability reveals discrepancies between the initial assessment of the issue and Google’s technical response.

Blaszczyk reported the flaw in May 2026. According to the researcher, the incident was initially categorized within Google’s vulnerability identification program as an issue with no security implications.

Despite this classification, development records indicate that a fix intended to address the vulnerable behavior had already been prepared by June.

The fix modifies how malformed messages are handled, preventing specific errors from causing an interruption to telephone service.

However, the existence of a fix within the Android codebase does not guarantee that it has reached all devices.

Android Authority examined the status of the fix and found conflicting information regarding the rollout schedule. While some references point to the October updates, others link the resolution to the November security cycle.

The GrapheneOS project also included a fix for this vulnerability in its early security releases, as documented in its official update history.

The distribution of the fix to Android devices depends on the manufacturers, the operating system versions, and their respective update schedules. What Android users should do...The primary protective measure is to keep the operating system up to date and install the security patches provided by the manufacturer.

Users can check their security update level in their phone's settings, though the exact location of this information varies depending on the manufacturer and the version of Android.

Security updates are particularly important because some vulnerabilities can be exploited without any user intervention. Even so, the presence of a recent update does not, by itself, guarantee that a specific patch has been installed.

The rollout of these updates may also vary between different models. As previously reported by TecheNet, some Google Pixel devices experienced issues following the September update—an incident that is separate and has no known connection to the vulnerability now being disclosed.

To date, the available information describes security tests and technical demonstrations. There has been no confirmation of an active exploitation campaign targeting Android users via this flaw.

However, the release of technical details underscores the importance of deploying the fixes, especially since the issue could affect a critical mobile phone function and disrupt emergency communications.

mundophone

 

TECH

DRAM crisis: Canon warns of more expensive cameras

After laptops, smartphones and game consoles, cameras are now also set to become significantly more expensive. Canon is warning of the effects of the ongoing DRAM crisis and plans to pass the higher costs on to consumers in the form of more expensive cameras.

The DRAM crisis, triggered by immense demand for RAM, SSDs and semiconductors for AI data centers, has already led to some extreme price increases for products such as graphics cards, gaming consoles and laptops over the past twelve months, with even the Nintendo Switch 2  already becoming more expensive. Next year, the crisis is also expected to hit cameras.

The global memory crisis, which has already impacted the market for computers, graphics cards, and other electronics, is now reaching the camera industry. Canon has warned that its equipment could become more expensive in 2027 due to rising production costs and difficulties in sourcing electronic components.

This warning was issued in an official Q&A document, released following the presentation of the company’s third-quarter 2026 financial results on October 27. In it, the Japanese manufacturer acknowledged that price pressure on memory components is expected to persist throughout the coming year.

This situation marks a shift in the company’s outlook. Previously, Canon had anticipated that memory costs would begin to decline in the second half of 2027; however, the manufacturer now expects no such reduction throughout the entire year.

Canon intends to pass these costs on to consumers. According to the company, a shortage of printed circuit boards and other electronic components is hindering the supply chain for its factories. To maintain production, Canon states that it is utilizing every available means to secure the necessary materials, even if it incurs additional expenses. The manufacturer explained that its strategy is to offset these higher costs through price increases starting in 2027, while carefully considering the potential impact on sales and profitability.

The effects are already visible in the company’s financial results. Canon revealed that the rising cost of memory, combined with other expenses linked to the situation in the Middle East, has reduced the expected operating margin for its imaging division in the fourth quarter by approximately five percentage points. The company also acknowledged that it has not yet been able to fully pass these additional costs on to its customers.

Artificial intelligence is driving the pressure on memory supplies. The issue is directly linked to the expansion of data centers dedicated to artificial intelligence, which are consuming increasingly large quantities of memory. Although cameras primarily use DRAM as temporary memory for processing images, videos, and sequences of photographs, its production relies on a supply chain that also serves computers, servers, and other equipment.

With semiconductor manufacturers prioritizing products intended for servers, the availability of components for mainstream devices remains limited. A study by TrendForce, published in September, estimated that contract prices for mainstream DRAM would increase by 10% to 15% in the fourth quarter of 2026.

The consultancy also forecasts that DRAM supply will remain under pressure in 2027, as the expansion of production capacity is not expected to keep pace with growing demand. These figures, however, pertain to the memory market and do not represent a price adjustment specifically announced for cameras.

Shortages are already affecting Canon’s camera production...The situation is also hindering the availability of certain equipment. In its financial report, Canon stated that demand for compact cameras remains high, but an insufficient supply of components is preventing production from keeping pace as expected. Despite this, revenue from compact cameras grew by more than 50% in the third quarter.

Meanwhile, sales of interchangeable-lens cameras totaled approximately 620,000 units—a 16% decline compared to the same period in 2025. The company has not yet disclosed which models will be subject to further price adjustments, nor the specific percentages or the markets that will be affected. Consequently, there is no confirmation yet as to how much more Canon cameras might cost in Brazil in 2027.

This warning, however, indicates that the effects of the memory crisis are spreading beyond the computer market, now impacting both professional and consumer photographic equipment.

These supply shortages typically raise costs, as seen across the flash memory space amid increased demand from AI data centers. Companies are bidding for a shrinking supply, driving prices up. Fujifilm alluded to this months ago, and now Canon is discussing it.

“For components that are in short supply, we are using every possible means to secure the quantities we need for next year. As a result, we expect higher procurement costs, but our policy is to offset those increases through price increases from next year onward,” Canon says.

“We have already been making pricing adjustments in response to cost increases as they occur, and we will continue to make decisions on passing on costs from a long-term perspective, while considering the potential impact on sales volumes and balancing that against profitability.”

So there are two really important things to pull out of this quote. First, Canon expects its costs to increase next year. And second, Canon’s modus operandi is to offset increased costs by passing them to the consumer, ergo, increasing product prices. It’s shrewd business, if the market will support it.

It’s not always possible to fully offset increased costs by raising prices. Sometimes markets won’t tolerate the necessary price increases, or a company may be willing to absorb some costs for its long-term prospects. For example, selling more cameras is good for Canon long-term, as it maintains more market share, sells more lenses, etc.

Canon says it will be a balancing act, but, notably, the company is coming right out and telling investors that yes, its policy is to pass increased costs to consumers. This makes perfect sense, but photographers need to hear the news and prepare accordingly.

Canon adds, in response to another question, that it doesn’t expect memory and raw material prices to decline in 2027, so this may be a long-term problem in the photography industry. Much like SSD and memory card prices have continued to increase, cameras may be the next victim.

Canon projects a significantly lower operating profit ratio in its Imaging business in Q4, which the company says is primarily due to higher memory costs and cost increases following the geopolitical strife in the Middle East.

“At this point, we have not been able to fully pass these higher costs on through pricing,” Canon says. “If these higher costs persist next year, we will absorb them through price increases.”

Canon is now warning of this in a Q&A document published alongside the announcement of its financial results for the third quarter of 2026. According to the document, it is becoming increasingly difficult to source electronic components such as mainboards in sufficient quantities, while the costs of sensors and DRAM, which is used as buffer memory in cameras, continue to rise. Canon says it is using "every possible means" to procure enough components for its cameras, but this entails higher costs than usual.

These higher costs are to be passed on to customers through higher prices, although Canon intends to take into account the impact of price increases on sales volumes and thus on profit. In simple terms, Canon is therefore considering absorbing some of the higher production costs itself, but price increases of varying magnitude are to be expected next year in any case. It is also worth noting that Canon had previously assumed that costs would fall again in the second half of  2027, but now believes that costs will continue to rise throughout 2027.

Memory costs have come up at every one of Canon’s earnings calls this year, so the topic itself isn’t new. What is new is the outlook: earlier this year, Canon said it expected memory prices to begin easing in the second half of 2027. Now, according to Canon’s Q3 Q&A summary, it doesn’t expect them to fall next year at all. The company says it has already secured the memory it needs for 2026, but expects to pay more for components in 2027, and that its policy is to offset those higher costs through price increases.

It’s worth noting that this isn’t entirely hypothetical. Canon told investors in July that it had already raised some prices in response to rising costs, including memory, though it didn’t say which products or markets were affected. It says it will continue to weigh further increases against their effect on sales.

Memory is central to how modern digital cameras work. High-speed continuous shooting, pre-capture modes, subject-tracking algorithms and more all lean on fast, plentiful memory. As AI-related demand has driven memory prices up across the industry, camera makers have found themselves competing for the same production capacity as data center operators, which are effectively the 800-pound gorillas in this market.

As a strategy, securing its memory supply early bought Canon time. But it isn’t a long-term solution if the underlying costs persist.

The wider picture...One company warning about costs is one thing. Several camera makers dealing with the same problem at once is another. Other manufacturers have also flagged rising memory costs to investors this year, and some have already started raising prices. That includes the US, where Blackmagic Design has raised prices on some of its cinema cameras and storage products, citing memory costs, and US buyers have already absorbed multiple rounds of tariff-driven price increases since last year.

When multiple camera makers are facing the same cost pressure, the probability that any one of them absorbs it entirely, rather than sharing it with buyers, drops considerably. Ultimately, this isn’t a Canon-specific story but an industry pricing story that happens to be told clearly through Canon’s investor relations disclosures.

Canon reports its full-year 2026 results early next year, and that call will be a useful checkpoint. By then, the company should have a clearer read on what it expects to pay for memory in 2027, and how much of that it plans to pass on to buyers. And because Canon sells more interchangeable-lens cameras than any other manufacturer, its pricing decisions are likely to be closely watched by the rest of the industry.

A Canon alert warns that camera prices could increase in 2027 if high memory costs and component shortages persist.

Key details from canon's warning:

• Cause of the crisis: High global demand for RAM, SSDs, and semiconductors from AI data centers has driven up the cost of DRAM—which cameras use as buffer memory—and mainboards.

• Timeline shift: Canon previously expected memory prices to drop, but now states elevated costs will likely continue into 2027.

• Pricing strategy: Because early stockpiling has run out, Canon told investors it will look to pass persistent production and component costs onto buyers if the squeeze continues.

• Industry-wide impact: Canon is not alone; other manufacturers like Sony and Nikon have also pointed to rising memory costs impacting their imaging divisions, and Blackmagic Design has already raised prices on certain cinema cameras and storage units.

by mundophone

DOSSIER

DIGITAL LIFE

What to consider if a company that made AI-related job cuts wants you back

Companies around the world are increasingly citing artificial intelligence when slashing jobs, fueling anxiety that a lot more workers may lose their livelihoods to machines.

Microsoft, Barclays Bank and supermarket chain Kroger are among the latest entries on a growing list of companies that have announced AI-related workforce reductions this year, including the elimination of jobs deemed redundant or shed to free up money for investing in the technology.

But the job cuts of the AI era, which are affecting tens of thousands of people, occasionally have an epilogue. Some companies, after discovering that AI models, applications and agents don't perform as well as expected, end up rehiring laid-off workers to fill eliminated positions.

It's unclear how common such reversals are, but they are happening often enough to acquire a nickname: the "AI boomerang."

"Two years ago, we said, 'Would this happen?' and now we're saying, 'Yes, it is happening in real time,'" said Lucy Beaumont, a global senior vice president at SHL, a U.K.-based technology company that develops employment-related tests and tools.

Hiring managers in fields including finance, marketing, law and health reported earlier this year that their companies had eliminated positions primarily because of AI or automation, only to hire again for the same or similar roles, according to recruitment and staffing firm Robert Half.

The most frequently cited reasons for restoring dropped jobs included the technology needing more human oversight than anticipated and its inability to replicate institutional knowledge, the firm said.

"When you need critical thinking, collaboration, creativity, institutional knowledge, those are things that AI just can't replicate," Bill Driscoll, a senior district president at Robert Half, said. "And I think that's where companies are starting to realize, 'Wow, we really do need people to do these jobs.'"

Klarna, the Swedish provider of buy now, pay later services, began recruiting human customer service representatives again last year. CEO Sebastian Siemiatkowski acknowledged in an interview with Bloomberg News that the company's reliance on an AI chatbot to cut costs had gone too far.

In June, Ford said it had hired or promoted 350 veteran engineers after AI tools the automaker deployed to do quality control work did not perform at the expected level.

"AI utilization has been an important part of our quality gains and that—along with deep technical expertise of our teams—is a powerful combination," a Ford spokesperson told The Associated Press.

However, companies don't always broadcast their U-turns on shifting resources or responsibilities from employees to AI systems, Beaumont said.

"Maybe their shareholders love that they did this right-sizing motion ... and then they have to turn around and quietly, with their tail between their legs, admit that they weren't ready for that or they shouldn't have done that," she said.

Management experts offered employment advice for workers as companies make greater use of AI tools and consider—or reconsider—their staffing needs while assessing the technology's capabilities.

Rehiring suggests the invisible work of humans is not easily replaced...When companies use AI to perform tasks previously done by people, they often miss the invisible work employees do beyond their formal job descriptions, said Betsy Summers, a principal analyst on the Future of Work team at market research and advisory firm Forrester.

That might include sensing emotions and moods, holding impromptu meetings and putting in extra effort when feeling inspired or otherwise motivated, Summers said.

"We don't know all the work that's required to do things well within the organization," she said. "The citizenship behaviors of team cohesion and team building and making sure that you have trust within your team, none of that is accounted for. So when you rip out humans who are doing that work and then expect agents to be able to then deliver exactly what humans have delivered, they are frequently mistaken."

Another factor that complicates assigning more work to AI-powered tools is that company procedures for tasks such as data processing and bookkeeping are not always well documented, experts said.

"Organizations are not spending enough time doing the up-front work to really truly understand workflows and where humans add value in the process," said Brooks Holtom, a professor of management at Georgetown University's McDonough School of Business.

In roles such as customer service, AI agents and chatbots frequently need to refer questions and complaints to a worker who can recognize emotions, assess the severity or complexity of a problem and respond creatively within established policies, Holtom said.

"It's that combination of creativity but staying in the guardrail that's really hard for a machine to figure out," he said.

Figure out what gives you a competitive edge as a human...As companies redirect resources toward AI, current employees and those asked to come back might want to emphasize skills that complement the technology, such as emotional intelligence and sound judgment, Holtom said.

"Really try to be able to articulate, what's your value proposition as a human?" Holtom said. "Don't be afraid to lean into the technology as well, and understand how to differentiate yourself from what machines do well. Embrace that, understand that, and you can become uniquely qualified to help with implementation."

Empathy is a human quality that technology has not fully mastered, Beaumont said.

"Think about an angry customer. They want to be listened to, they want to be heard, they really want a human on the other end of the line," Beaumont said. "Can an AI replace that right now in a way that feels meaningful to individuals? Perhaps not."

Know your leverage and use caution if you're asked back...Many people are understandably reluctant to return to a company that laid them off as part of a reorganization around artificial intelligence, according to Beaumont.

"To come back to a business where you've been let go, replaced by AI, to be asked back to that business, your trust is gone, your psychological safety there, your relationship with the employer, and that's very, very difficult to build back," she said.

Workers offered their old jobs back are asking for more money and to know how AI might impact future staffing, Beaumont said. Requesting additional training and a clear explanation of your opportunities for advancement could be part of the discussion, she said.

Communicate ways you would contribute value as a returning employee, outlining a new job description "that leverages the technology but also your expertise," Holtom said. You could end up in a higher-ranking position that pays more than the one you left, he said.

Before accepting an offer, seek a commitment such as a one-year renewable contract that includes a buyout clause if the job is terminated in less than a year, Holtom suggested.

Don't discount the distinctly human values of motivation and curiosity, which make people efficient and effective, Summers said.

"Within our societies, work serves a broader purpose than just creating value for shareholders or customers or constituents," she added. "Work serves a human purpose as well. It's the manifestation of our search for meaning and purpose. We want to contribute to our communities as human beings."

If an employer previously eliminated your role in favor of automation and is now asking you to return, you are part of a growing trend. Many companies are experiencing "rehire regret" as automated systems underperform, meaning you hold significant leverage. Prominent brands like Klarna and Ford have had to rehire or promote human teams after finding that chatbots and automated quality control couldn't match human expertise.

If you are negotiating a return, consider these critical factors regarding your value, your leverage, and your contract stability:

1. Leverage your unique human value...Firms often adopt AI models before fully understanding how much of their workflow depends on human intuition and judgment.

• Define your value proposition: Explicitly state what you bring to the table that machines cannot replicate, such as advanced emotional intelligence, ethical decision-making, and deep contextual experience.

• Emphasize empathy: For customer-facing roles, remind them that frustrated users require authentic human connection, active listening, and problem-solving that AI tools still struggle to meaningfully deliver.

• Position yourself as the "final line of defense": Frame your return around quality control—auditing AI outputs, fixing its systematic errors, and managing tool implementation.

2. Renegotiate your scope and compensation...Because the company is likely re-entering negotiations "with their tail between their legs", do not simply accept your old job back on the exact same terms.

• Propose a higher-ranking role: Work with management to co-create a new, elevated job description that combines your domain expertise with technology management.

• Demand a raise: Restaffing an disrupted department is incredibly expensive for an enterprise; use this premium to request higher compensation than what you earned before you left.

• Secure training commitments: Request that the company explicitly fund your upskilling and continuous training so you can effectively integrate and oversee new automated workflows.

3. Build in strict financial protections...Past behavior remains one of the strongest indicators of future corporate actions. If leadership rushed to cut costs once, they may attempt it again when a newer iteration of software debuts.

• Request a guaranteed contract: Try to secure a one-year renewable contract rather than standard at-will employment.

• Include a severance/buyout clause: Ensure the contract states that if your position is terminated again within a specific window (e.g., less than a year), the company owes you a substantial, predefined financial payout.

• Demand transparency: Ask flat out how AI tools are expected to impact broader department staffing over the next 18 to 24 months

If an employer that laid you off to replace you with artificial intelligence asks you to return, you hold the leverage, but you must evaluate the company's technical stability, negotiate strict job security safeguards, and demand a significant premium on your compensation.

A growing wave of companies—ranging from tech startups to legacy automakers like Ford and fintech giants like Klarna—are quietly walking back aggressive AI automation plans. Research indicates that more than half of employers that cut staff in favor of AI now regret it, realizing they handed critical processes to models before fully understanding how much those tasks relied on human oversight, sound judgment, and empathy.

If your former company reaches out with an offer to return, keep the following strategic considerations in mind:

1. Assess why the AI strategy failed...Before accepting, you need to understand exactly what went wrong during your absence so you can prevent history from repeating itself.

• Determine the operational gap: Did the AI tools fail to perform at the expected quality level (as seen with automakers deploying automated quality control)? Or did customer satisfaction tank because clients demanded human interaction?

• Identify your exact role: Are they asking you back to do your old job, or are they bringing you back as a "fixer" to clean up the errors and hallucinations the AI generated?

• Gauge management's self-awareness: Is leadership transparently admitting they overreached? Be wary of management teams that try to minimize their previous strategic errors, as they may try to automate your role again the moment a new software update drops.

2. Define your value proposition as a human...When discussing a return, do not simply slot back into your old, pre-layoff routine. Position yourself as a strategic partner who manages and elevates the technology.

• Emphasize complementary skills: Clearly articulate the human-centric skills you bring that software lacks, such as complex problem-solving, deep technical expertise, emotional intelligence, and team cohesion.

• Rewrite your job description: Collaborate on a new role that explicitly balances your expertise with the company’s AI infrastructure. You should be positioned to oversee, train, or audit the AI, effectively future-proofing your position.

3. Demand financial and contractual protections...Because the company already proved it views your role as a line-item cost that can be automated, you must protect yourself against future tech-driven "right-sizing" motions.

• Ask for a premium: Restaffing an disrupted department is incredibly expensive for an employer. Because they need your institutional knowledge immediately to correct ongoing mistakes, you have the upper hand to demand higher pay or a higher-ranking title than the one you left.

• Negotiate a buyout clause: Workplace experts suggest securing a guaranteed contract (such as a one-year renewable agreement) that includes a strict severance or buyout clause if the job is eliminated due to technology or restructuring in less than a year.

• Verify employee benefits: Some companies funding heavy AI infrastructure have quietly cut 401(k) matches or reallocated salary raises to fund software licenses. Audit the company’s current benefit structure to ensure you aren't walking back into diminished total compensation

mundophone

 

TECH

Mystery of 60-second fluctuations in power grids solved

Why do our grids show a brief spike in electricity demand every minute? A team led by Dirk Witthaut from Forschungszentrum Jülich has now demonstrated that this is due to regular power checks carried out by large-scale battery storage systems. These findings could help to make our electricity supply more robust, as the team reports in an article published in Nature Energy.

Tennet Germany, the largest transmission system operator in Germany, has observed over several years that the frequency in the European power grid dips briefly every 60 seconds. While these deviations were not critical to grid security, they were clearly measurable and unusually regular. The cause of this phenomenon was initially unclear.

A team from Jülich led by Dirk Witthaut, together with researchers from the University of Cologne and the Karlsruhe Institute of Technology, set out to investigate – and found the answer: the cause appears to be the many large-scale battery storage systems that check their state of charge every minute. This leads to measurable fluctuations in electricity consumption, which can contribute to turbine wear and tear, for example. The battery storage systems ensure an uninterrupted power supply, and their numbers have increased significantly worldwide in recent years. The researchers found that this was not a European phenomenon, but a global one.

Many devices, a common rhythm...“It really was years of detective work,” says Witthaut. The grid frequency pattern is shaped by many different factors. For example, it reveals the influence of electricity trading, with power stations being switched on and off abruptly, as well as the fluctuating feed-in from renewable energy sources, and various other disturbances.

“It quickly became clear to us that the cause must be something controlled by the clock,” says Witthaut. There was no other way to explain the precise 60-second cycle. It was also clear that: “The observed effect can only be produced if many devices are consuming electricity at the same time.” A single consumer or generator could therefore be ruled out.

Measurements on campus provide the clue...But how do all these devices synchronize their consumption? The researchers finally found the crucial clues with the help of the Living Lab Energy Campus. This research infrastructure on the Jülich campus is primarily used to develop and test highly integrated energy supply systems. The team installed measuring points on end devices at a total of 80 locations – and sure enough, the same periodic pattern in electricity consumption was observed precisely at the locations where large-scale battery storage systems are designed to ensure an uninterrupted power supply. “We took a closer look at one of them and found a spike in electricity consumption at the start of every minute due to the power check,” explains Witthaut.

As there are no binding standards for such power checks, Witthaut suspects that most battery manufacturers assumed that a 60-second interval would be sensible. The start of every minute was simply chosen as the test time. “It is entirely possible, however, that battery storage systems are only part of the problem here,” says Witthaut. “It cannot be ruled out that other digitally controlled appliances – such as air-conditioning systems, heat pumps, washing machines, and dishwashers – may also exhibit similar patterns.”

Less synchronization, less wear and tear...The periodic fluctuations in electricity demand are already having an impact on hydroelectric power stations. The brief spikes in demand mean that the turbines slow down momentarily before accelerating back to their original rotational speed. This increases wear and tear. With the expansion of storage technologies, such problems could become even more prevalent. In Ireland, for example, where data centres with large-scale battery storage systems currently account for around a fifth of the country’s electricity demand, the amplitude of the 1-minute signal quadrupled between 2016 and 2021. The solution, however, would be simple: “You would simply have to distribute the checks randomly, rather than always starting them at the same time,” says Witthaut.

One fallen power line exposed a growing AI data center problem — here’s how to fix it...A power line went down outside of Washington, DC, this week. Normally, the grid would only need a few seconds to recover from such an event. But this one took more than 10 minutes because data centers stopped drawing over 3 gigawatts’ worth of power nearly simultaneously.

The event caused voltage across the PJM grid to spike from Northern Virginia to Chicago, according to data collected by Ting Labs, a startup that runs an IoT sensor network out of people’s electrical sockets.

The event didn’t cause a blackout, but it did cause lights across the region to flicker. The incident demonstrated the effect that data centers can have on the grid — an outcome that experts believe will become more frequent.

Northern Virginia, which is in PJM’s territory, is home to the highest concentration of data centers in the world.

“It’s the canary in the coal mine,” Ricardo de Azevedo, CTO at ON.Energy, told TechCrunch. These sorts of events involving large loads like data centers are “happening more and more,” he added.

The event echoes one that happened two years ago, also on PJM’s grid, and it could foreshadow larger events if data centers aren’t built to more elegantly handle disruptions to power supplies. The PJM Interconnection manages grids from New Jersey to Illinois and serves 67 million customers, making it the largest grid operator in the United States.

When the power line went down this week, it triggered data centers to switch to backup power, and about 3.1 gigawatts of load vanished in about 30 seconds, according to PJM data. The grid appeared to recover somewhat, but a short time later additional loads dropped off. At its peak, PJM’s grid had an extra 3.49 gigawatts of electricity on it. It took another 11 minutes before it stabilized. The disconnected data centers represented around 3% of total demand on PJM at the time, according to Reuters.

A few percent may not sound like much, but the electrical grid needs to operate in a state of near-perfect balance, with supply and demand closely matched. If they don’t, voltages can sag or spike. The grid and devices connected to it can tolerate small fluctuations, but if those fluctuations grow too large, they’ll trigger fail-safes within the grid or within individual facilities, causing them to disconnect. 

When data centers in Northern Virginia sensed the fluctuation caused by the failed power line, they switched to backup power, which removed their load from the grid. As more data centers made the switch, they removed greater amounts of load from the grid. What started as a relatively small drop in supply became an even larger drop in demand, sending supply surging and causing light bulbs to flicker. 

Large-scale battery energy storage systems are the source of mysterious 60-second fluctuations detected in power grids worldwide. An international study led by researchers at Forschungszentrum Jülich, published in the journal *Nature Energy*, has solved the puzzle that had been baffling transmission system operators across the globe.

European regulators recently published a nearly 500-page report on the outage, attributing it to numerous and interacting factors. The report cited work from researchers in WSU’s School of Electrical Engineering and Computer Science regarding poorly understood frequency oscillations throughout the power grid that were part of the event. The work cited came from PhD students, Arash Nezam Sarmadi, Tianying “Lily” Wu, Yuan Zhi, and Salman Shiuab. All have since graduated.

“We have the best students, and it is really something to be proud of for WSU that this report, originating in Europe, cited publications from four of our students, and that these tools we developed are being used — both in real time and for offline analysis of complex events — for this Iberian blackout,” said Mani Venkatasubramanian, Boeing Distinguished Professor in Electrical Engineering who led the work.

It was a beautiful spring day last year when the European power grid began oscillating in frequency. Venkatasubramanian, director of WSU’s Energy Systems Innovation Center, who holds a joint appointment at Pacific Northwest National Laboratory, has worked with graduate students for decades to develop online monitoring tools to detect these harmful oscillations that cause fluctuations in power outputs in the power grid. The work was funded by the U.S. Department of Energy and by the Power Systems Engineering Research Center. He received a patent and started a company, Phasor Informatics LLC, that has licensed the technology. The software uses a suite of algorithms to assess oscillation problems in transmission systems and to notify operators when they occur.

Two five-minute oscillation events occurred that morning in Spain, and each time, power grid operators in Paris, France, received notifications from the software developed at WSU. The operators in Spain, in coordination with their counterparts in France, took actions to dampen them.

“The phenomenon that a forced oscillation resonated with the system mode was something that we had postulated in our papers, and we showed the theory on how that would affect the large power system,” said Venkatasubramanian. “So those were what caused the big oscillations that were seen between Spain and France and possibly other countries.”

As grid operators reacted, however, voltages on the power lines increased above acceptable limits, and power plants automatically disconnected in a cascading event. In less than two minutes around noon, about 60 million people across Spain and Portugal lost power in an outage that lasted 12 to 16 hours.

A new open-source tool developed at WSU helped Iberian blackout investigators uncover the likely source of the first oscillation that originated in Spain.

Using a new open-source tool developed at WSU that uses measurements with a slower reporting rate to detect faster oscillations, the Iberian blackout investigators helped uncover the likely source of the first oscillation that originated in Spain.

“That was exciting because we developed this tool just two years ago,” Venkatasubramanian said.

Oscillations remain a challenge for the power grid. The researchers have developed controls intended to dampen them, but they don’t fully understand why those actions don’t always work.

The U.S. has strict regulations and a more robust system than many other countries, but there have been oscillation events here, which are warnings for the system. In recent years, power grid operators have been planning for and seeing significant changes in the grid’s management. The cost of renewable energy production has dropped dramatically so that solar and wind power are now the cheapest and often the most attractive ways to produce electricity. Renewables provide variable amounts of energy during the day and are often distributed around a region instead of being centralized. At the same time, climate change, extreme weather events, and the growth of AI and data centers add to power grid challenges. Meanwhile, the system remains hampered by the lack of transmission support.

“Because of economic and societal issues, it’s very difficult to build new transmission lines in the U.S., and that is stressing the system more and more,” said Venkatasubramanian. “It is requiring us to become smart in how we use the existing resources to serve this growing demand.”

Venkatasubramanian’s group is developing open-source planning tools for modeling and analyzing the complexities and uncertainties that come about from the use of renewables and from extreme weather events.

The mechanism behind the phenomenon:

• Involuntary synchronization: Thousands of commercial battery systems around the planet have been programmed by their manufacturers to perform internal state-of-charge tests at the very beginning of every minute.

• Instantaneous consumption spike: While the impact of a single battery is negligible, the simultaneous and precisely coordinated behavior of a vast number of units creates a measurable fluctuation in the frequency of the global power grid.

• Lack of standards: The 60-second pattern emerged spontaneously because manufacturers simply chose an intuitive time interval, with no regulatory standard in place to account for this domino effect.

Practical consequences:

• Infrastructure wear and tear: These brief, cyclical oscillations force hydroelectric turbines to momentarily slow down and then speed back up, increasing the mechanical wear on the equipment.

• Amplification through technological advancement: As the energy transition progresses and more storage systems are connected to the grid, the amplitude of this signal grows. In Ireland, driven by the expansion of data centers equipped with battery backups, the magnitude of this impact on the grid has quadrupled over a five-year period.

The proposed solution... The definitive resolution to the problem is surprisingly simple. The study suggests that the issue can be resolved by decentralizing the tests—specifically, by introducing a randomized time distribution for when the batteries perform their checks. Instead of responding to the clock in an identical manner at the very first second of each minute, the devices would take their measurements at staggered times, completely neutralizing the collective peak load.

mundophone

I need your help via Pix (QR code above) or using the code