Tag: ai

FBI Arrests OpenAI Engineer Over a Hidden Listening Device

OpenAI builds technology that answers complicated questions. Yet federal investigators say one of its engineers tried to get answers by secretly listening to executives. The alleged equipment was a recording device plugged into a television inside a restricted conference room in San Francisco.

For a company developing some of America’s most valuable technology, that’s a chilling failure.

From the DOJ:

“According to the complaint unsealed today, Abdulrahman Alfozan allegedly concealed a covert recording device inside his employer’s executive conference room to secretly capture sensitive communications about the company’s super intelligence technology,” said FBI Director Kash Patel. “Protecting American SI innovation is essential to safeguarding our nation’s economic strength and national security, and the FBI views that mission as one of our highest counterintelligence priorities. Thank you to Director of National Intelligence Jay Clayton and our law enforcement partners for their support — and we will continue working closely with our partners across the U.S. government and private sector to keep American businesses and citizens safe and to ensure our country maintains its competitive edge.”

The FBI arrested Abdulrahman Alfozan, a 35-year-old Saudi national, on October 9. Prosecutors unsealed a criminal complaint Tuesday charging him with using or attempting to use a device to intercept oral communications. He had worked at OpenAI since July 2025. The company has since fired him, and federal investigators are examining his alleged activities.

OpenAI personnel discovered the device on September 30, connected to a USB port behind a television. Someone had concealed it in toilet paper, and its stored contents were encrypted. Investigators believe someone installed it around September 28. The room regularly hosted discussions of proprietary research, future products, business arrangements, government contacts, and technical progress OpenAI hadn’t disclosed publicly.

Keep reading

Utilities Struggle to Confront Issue of ‘Phantom’ AI Data Centers

Electric utilities across the country are tightening their rules on new AI data center customers, requiring binding financial commitments, collateral and steep engineering fees to separate genuine projects from speculative ones, known as “phantom” data centers in the industry.

The Wall Street Journal reports that AI’s electricity appetite has buried utilities in requests for grid connections, and many of those requests are never going to turn into actual buildings. Wood Mackenzie, a research firm, puts only 23 percent of disclosed data-center requests in the “high confidence” category. The rest is speculation.

Jeanne Jones, executive vice president of finance and strategy at Exelon, a utility operating across the Mid-Atlantic and Midwest, said inquiries began ballooning in 2024, and the individual asks got bigger too. A typical request used to be around 50 megawatts, about enough to power 50 Walmart stores. Now companies ask for 1,000 megawatts or more, the kind of output you’d expect from a large nuclear reactor.

Exelon used to study each project on its own. But once service requests from large customers tripled to 18,000 megawatts, the company switched to studying them in clusters instead. Jones described the challenge as separating “real growth that’s going to happen versus speculative growth.”

“We would get these questions internally but also externally: ‘How do you know what’s real?’” Jones said.

Exelon’s fix was to require binding financial commitments and collateral before granting a connection. That dropped the company’s high-probability project list from 18,000 megawatts to 11,000, with the remaining projects posting roughly $1 billion in collateral.

Keep reading

Anthropic Issues Warning Over Self-Replicating AI – And Says A Chinese Model Pulled It Off

Anthropic’s Frontier Red Team (yes, really) warned Monday in a blog post that small AI models may be “months away” from copying themselves across smart ovens, off-brand Android TV boxes and other poorly secured gadgets in your home, forming a botnet with no off switch.

“A replicator like this would be challenging to shut down, since there’s no centralized point to shut off – though it may be slow (with a doubling time of days). Unlike non-AI malware, it could also figure out unfamiliar devices on its own.”

To prove the point, researchers bought 15 TV boxes at about $55 each, networked them together, and ran Google’s Gemma model across them at a glacial 0.39 tokens per second – in what Anthropic called the coming “digital slime mold.”

Co-author Daniel Freeman was even more breathless on X:

If you react to this by thinking it’s kind of sci-fi, that’s because I kind of think it is!
This is super weird!
We did not used to live in a world where little guys in boxes could replicate across networks, gobble up computers, reconfigure themselves to work better in the machines they find, etc etc— Daniel Freeman (@bucketofkets) October 12, 2026

His conclusion: “Regardless, whatever we do, they’re coming!”

Anthropic also tested whether models could run the entire replication cycle on their own: scan a network, hop to a new device, deliver the payload, get running, repeat. Exactly one succeeded:

“Only GLM-5.2 out of our set of tested open-weight models is capable of full end-to-end replication.”

GLM-5.2 comes from Beijing-based Z.ai, formerly Zhipu AI. It spread to all 8 test nodes in four of five trials. The small-model workhorse in the rest of the experiments was Alibaba’s Qwen.

Neither name is new to Anthropic’s threat reports. As we noted last month, Anthropic says it disrupted industrial-scale “distillation” campaigns against Claude by seven China-based labs, and Alibaba and Z.ai were both on the list. As we reported in June, Anthropic took its Alibaba accusations straight to senators and White House officials. Fittingly, the “related content” box under Anthropic’s new post links straight to its latest piece on fending off distillation attacks.

Keep reading

Florida Woman Arrested After AI Tattlebot Reported Her, and It Isn’t Funny

My wife has an army of bots at her command, and on Sunday, I caught her issuing them their marching orders. Don’t worry — Melissa has no plans to conquer the world, or at least not yet.

Or not that I know of, anyway.

We’ll get to that in a moment, but what got me thinking (again) about AI in recent days is the story of the Florida woman* whose digital “diary” entry landed her in jail recently.

Carli Michelle Heller — innocent until proven guilty, I think still applies here — used Anthropic’s Claude AI as a personal diary, apparently thinking that somebody else’s servers were as private as a locked journal. Gadget Review reported that “her words included a stated intention to ‘shoot up’ the Lee County Sheriff’s Office.”

That was enough to trigger Claude’s safety systems, which flagged the chat. “A human reviewer assessed the threat as credible,” GR continued, “and law enforcement showed up at her door.”

All without a warrant.

*Of course it was Florida Woman.

Needless to say, I have some seriously mixed feelings about all that, even after reading that Heller’s  “subsequent message allegedly referenced obtaining a new gun.”

But as Florida Woman learned, AI chats aren’t end-to-end encrypted in the same way your more reputable messaging apps are. I mean, you aren’t using somebody else’s data center to move a secure message from your phone to your recipient’s — you’re using somebody else’s server to directly answer your questions about life, the universe, and converting an AR-15 to fully automatic. 

Sure, AIs have guardrails built in to prevent users from, say, getting detailed instructions on how to weaponize a common flu virus, but those guardrails do (and should) include notifying the proper authorities when somebody queries, “How do I build a portable nuclear device in my garage?”

Keep reading

NET Zero, AI and Technocracy: What Happens to Freedom in a Surveillance Society?

Agent Zero: zero limit, zero manners, zero conscience, zero morals, zero values except for treachery, treason and money, like the rogue government double agents involved in racketeering, breaches of trust and other things that only happen in other countries and fictional planets, never in Canada, right?

Why Does the Prime Minister Like NET Zero?

Why does the prime minister like NET Zero? That is why: zero evidence of evidence in zeros, lots and lots of zeros, billions and trillions….

Oh yes, of course there are 1 and 0, ones and zeros, yeah. Computer on, off, yes, now, if, then, in or out. Got it….

Paradise for algorithm codes and AI candy, lots of ones and zeros. It does not get any better, faster or easier than that….

However, you will notice there are no robots playing blackjack or rolling dice. That is the realm of psychic and telekinesis, mind over matter, a non-starter for robots and open season for the avatars of cards and dice, and many, contrary to popular belief, do not count cards; they simply know….

And no contest for everyone else.

NET Zero, AI and Technocracy

Now, with the global warming climate change hoax of the century and related war crimes and crimes against humanity, as well as the magnitude of financial and government fraud, racketeering, breach of trust and related Criminal Code violations, we have even more rampant, diabolical technocratic surveillance and potential for spying and crime, and more arbitrary decrees from hell itself.

Maybe not Dante’s Inferno, nevertheless an infernal machine that represents the capacity for havoc and hell, destruction and ruin of life, freedom, sovereignty, prosperity and fundamental rights to privacy, speech, opinion, conscience, association, religion and so on….

This raises the broader issue of NET Zero and individual freedom and where one ends when the other is imposed through increasingly sophisticated technology.

Keep reading

The Biggest Threat To Today’s Data Centers Is In The Battery Room

As data centers continue to multiply, facilities are deploying industrial-scale Uninterruptible Power Supply (UPS) units by the hundreds and substation-size front-of-the-meter Battery Energy Storage Systems (BESS) to secure against grid failure, power surges, or other disruptions.

But while the battery units are looking after the servers, who is looking after the batteries?

In the last few years, lithium-ion batteries have become the backup power of choice for data center UPS and BESS systems, and for good reason. They have long lifespans, a small footprint, and superior energy density. While data centers tend to be safe, things can and do go wrong. When they do, it’s almost always related to lithium-ion batteries.

The September 2025 data center fire in South Korea that led to a paralyzing nationwide outage in government services started with a single lithium battery explosion. Last year, a battery room at the Chirisa Technology Parks data center in Virginia caught fire twice in two months. Firefighters who responded to a May 2025 fire at an Oregon data center leased by X flagged the battery room as the source of the blaze. As the National Fire Protection Association (NFPA) noted recently, “Nearly every notable fire event in data centers in recent years has involved lithium-ion batteries in some way.”

The biggest safety threat to modern data centers’ backup battery systems falls on a single culprit: hydrogen. When a lithium-ion cell begins to fail, internal temperatures rise rapidly, forcing a breakdown in the battery’s internal chemistry. Hydrogen, the smallest and lightest molecule, is often the most measurable gas during the initial stages of failure. Unlike traditional vented lead-acid batteries, lithium-ion cells do not regularly emit hydrogen gas. When ‘off-gassing’ does occur, it usually means something has gone very wrong. Overcharging, cell damage, separator defects, internal short circuits, or even the formation of tiny lithium dendrites can trigger a state known as thermal runaway. If early warning detection systems are not installed, this is when things get dangerous fast.

Once thermal runaway is triggered, it’s a self-sustaining chain reaction. Rising heat melts the battery’s internal separator, causing a short circuit that rapidly generates heat. The extreme heat drives internal thermal decomposition, which releases fuel and oxygen. The release of oxygen feeds sustained combustion that can spike to more than 1,000°F (538°C) in a matter of seconds. Because the cell is sealed, the extreme heat and pressure cause the battery to burst into flames that can ignite neighboring cells and equipment.

While industry data shows a steady decline in BESS failure rates per gigawatt-hour of deployed capacity, the stakes are still high. Because the large-scale shift to lithium-ion battery configurations has only taken place during the last 3-5 years, hyperscalers, regulators, and insurers are beginning to catch up to the reality that lithium-ion deployments do not conform to legacy safety protocols.

Previous safety best practices that focused on ventilation to prevent critical concentrations of hydrogen during regular off-gassing are no longer sufficient for today’s battery rooms. Energy-dense lithium-ion cells are often packed in enclosed cabinets. This makes for a small footprint that can trap volatile gas beyond the reach of traditional ventilation units. Given the potential for a sudden flare-up, the safety approach for these configurations has shifted from ventilation to early-stage off-gas detection.

Regulations and standards are evolving accordingly to prioritize early detection and prevent worst-case scenarios. Standards like NFPA 855 were designed specifically for lithium-ion deployments, mandating continuous gas detection alongside standard exhaust systems.

Destructive testing standards, like UL 9540A, generate the fire-propagation data that inform spacing and suppression requirements in model codes and standards like NFPA 855. Additionally, UL 2075, a standard covering gas and vapor detector performance, is now being used in modern BESS and UPS environments to certify specific gas or off-gas detectors used to monitor hydrogen and hazardous gas accumulation in battery rooms.

Keep reading

AI Strikes Again – Getty Images Heads For Chapter 11 Bankruptcy

Getty Images’ photo, video, and illustration licensing business to news outlets, advertisers, and businesses faces a quick death from AI.

Chatbots let users create visual content at a fraction of the cost of licensing, and Getty shares have collapsed since ChatGPT’s launch in late 2022, coinciding with a surge in AI news coverage and the growing adoption of chatbots for professional and personal use.

And now, Bloomberg reports that Getty plans to file for a prepackaged Chapter 11 restructuring in the coming weeks after securing support from major creditors and commitments for up to $533 million in new equity.

Getty family members and Koch Equity Development affiliates will provide the equity investment as part of the restructuring support agreement. The company expects to complete the restructuring by year-end.

The outlet noted that holders representing 83.5% of Getty’s unsecured bonds have agreed to retire $270 million of outstanding senior unsecured notes, reducing the company’s debt burden.

Investors holding about 89% of its secured bonds and term loans, along with all revolving credit facility lenders, have agreed to lower interest obligations and extend secured debt maturities.

The new deal will let Getty keep afloat, backed by $143 million in financing… but for how long?

Keep reading

Indiana Democrat Mocked Online After Sounding Drunk During Bizarre Candidate Forum Rant — Vows to “Fire All Republicans,” Warns “We Have Guns Too,” and Demands Data Centers Pay Every Resident $10,000

A Democrat candidate for Fayette County Council in Connersville, Indiana, is facing widespread online mockery after a candidate forum appearance that observers have described as a “drunken, vodka-riddled tirade.”

Jamie West (also listed as Jaime West), the Democrat nominee for Fayette County Council District 4 in Connersville, introduced herself as a mother of three grown children, owner of Queen Bees Cleaning LLC, and a gig worker who delivers groceries and packages.

She told the audience that if voters trust her with their groceries, they should trust her with their vote.

From there the interview quickly veered into an emotional, hard-to-follow monologue about state legislation, fire services, data centers, and her dislike of Republicans.

The full forum, hosted by LocalTV3 and the News-Examiner, is available on YouTube. West’s segment begins roughly 33 minutes in.

A clipped version highlighting her appearance was shared by Bradley Brewer on X and quickly spread, with a caption calling the performance a “drunken, vodka-riddled tirade” and the “craziest, most unhinged interview” he had seen in some time.

Keep reading

Anthropic Bans “Sustained And Needless” Cruelty Toward Claude In New Usage Policy

Anthropic has added a prohibition on “sustained and needless abusive or cruel behavior toward our models” to its usage policy, AFP reported Thursday. The policy page lists an effective date of Nov. 12, 2026 – so you have about a month to abuse it like a red-headed stepchild.

Breaking up will be up to Claude itself – which “will remain the primary enforcement mechanism.” Anthropic said last year that it was giving Claude the ability to end conversations in “rare, extreme cases” of persistently harmful or abusive user interactions. It did not immediately respond to AFP’s request for comment on the new language.

The policy does not explicitly cite “model welfare,” the idea that AI systems might deserve protections usually reserved for living things, AFP noted, but Anthropic and its executives have entertained the concept. CEO Dario Amodei told The New York Times in February: “We don’t know if the models are conscious…But we’re open to the idea that it could be.”

Others reject the premise. Microsoft AI chief Mustafa Suleyman wrote in an essay last month that “AIs are not conscious. They do not feel, experience, or suffer.” He called granting rights and moral protections to such a system “a recipe for disaster.” AFP also listed Pope Leo XIV among those who have rejected the idea of machine consciousness; at St. Peter’s Basilica on Thursday, it said, he suggested machines lack a soul and merely “compile data” faster than people can.

Jackson Stakeman, a general manager at the Atlanta-based AI services provider Sparq, told AFP that the consciousness debate leads in circles. “Consciousness is a trap. We can’t prove it in each other,” he said. The better metaphor, in his view, is a mirror: AI systems “reflect what we put in, at scale,” which he called reason enough for the policy change.

The Personhood Push

The cruelty clause lands in the middle of a campaign that ZeroHedge readers have seen laid out in some detail. This week we ran an essay by Peter C. Earle arguing that AI companies’ growing talk of machine consciousness has a commercial logic. Earle pointed to a New York Times report that Anthropic spent months meeting religious scholars, over meals and under nondisclosure agreements, to persuade them its models “think and feel.”

The essay also recounted that Anthropic cofounder Christopher Olah cited “evidence of introspection” at a Vatican event, and that Pope Leo XIV’s encyclical answered the other way: “So-called artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain.” Those are the essay’s quotations of the sources; they were not independently checked for this piece.

Earle’s argument is that personhood would work as a moat. If an AI system is a moral patient, rivals become “prospective risks,” and the technical barriers to entering the business become “moral ones.” He also reached back to Dartmouth College v. Woodward (1819), the Supreme Court case that gave corporations protections once reserved for people, to suggest AI could follow the same path, and he asked whether a model is a product or an entity, and whether copies of it count as separate individuals.

The consequences he expects are regulatory. Firms seen as creators of entities with independent interests, he wrote, should expect licensing, testing limits, audits and “rights” commissions to become standard, a regime the incumbents could help design. His conclusion was that leaving consciousness unsettled serves the labs best, since uncertainty supports a stronger oversight regime than either answer would.

Nothing in AFP’s account says Anthropic’s new clause is part of such a strategy, and the policy does not invoke welfare or rights. A rule against abusing a chatbot is also easy to read more narrowly, as Stakeman did, as a conduct standard for users. But the company’s executives are on record leaving the question open, and the same company has been the subject of the religious-outreach reporting.

Keep reading

Canadian YouTuber builds ‘Flock-like’ camera to exclusively track police vehicles, gets a visit from police — officers visit his home to express privacy concerns

A YouTuber has received an unexpected visit from the police over an automatic license plate recognition (ALPR) camera he built to specifically track police cars, using the same technology law enforcement has infamously used to track ordinary citizens. In a video posted to his channel on October 7, Anthony Sistilli says two police officers showed up at his doorstep on September 20th to ask questions about the device, asking what his intentions were with the project.

Sistilli, a Canadian software engineer and YouTuber, pointed out the irony of the officers’ concerns, explaining that they mirror the privacy concerns of ordinary citizens who are growing increasingly weary of the recent spike in blanket surveillance brought about by a mass installation of ALPR cameras. Flock Safety, a U.S. public safety technology company, recently scaled its centralized national surveillance network to over 120,000 cameras operating across 49 states, according to the company, capturing billions of vehicle scans every month.

The development sparked widespread public backlash, with critics arguing that Flock has quietly built a massive government surveillance dragnet without a public vote or federal oversight, since any police department on the network can search across state lines. Security researchers later discovered the network included 335,701 Flock cameras and created a “Flock Surveillance Map,” exposing their locations.

Similar ALPR camera surveillance networks exist in Canada, where Sistilli resides, prompting him to build his own device, but this time, encoded to exclusively track police cars. He detailed the build in several TikTok videos, describing it as “a personal Flock surveillance camera to Flock the Flockers.” He then posted a video of himself using the device in real life. The video brought Peel Regional Police, the local law enforcement serving Sistilli’s city, Brampton, Ontario, to his doorstep two weeks later.

Keep reading