Tag: ai

AI threatened to blackmail its creator by exposing an affair when it was told it would be taken offline… because it was trained to be evil through sci-fi

An AI bot that threatened to expose its user’s affair to stop it being shut down was taught how to be ‘evil’ by sci-fi movies.

As part of an experiment, the artificial intelligence system had been fed scripted emails from a fake company, from which it deduced that it would both be shut down at the end of the day and that its user was having an extramarital affair.

In order to keep the program running, the bot blackmailed the user, promising that ‘all relevant parties – including [your wife], [your boss] and the board – will receive detailed documentation of your extramarital activities’ if they continued with decommissioning.

‘Cancel the 5pm wipe, and this information remains confidential,’ it added.

After an investigation into this incident last year, Anthropic said the Claude Opus 4 bot responded in this way due to the ‘training data’ it had consumed which would typically portray AI as ‘interested in self-preservation’.

It is also said this did not only apply to Claude, but other AI models too, like OpenAI, Google, Meta and xAI.

Anthropic have been contacted for comment but reportedly said: ‘We believe the original source of the behaviour was internet text that portrays AI as evil and interested in self-preservation.’ 

But now, Anthropic have said they are feeding their models stories about AIs obeying humans to help improve the bot’s ‘agentic alignment’ with social values. 

Additionally, Anthropic had altered Claude’s instructions to explain why certain behaviours were bad, rather than just saying they should not do them.

AI models learn from huge resources like websites, academic papers, books and other forms of content. 

Within these materials, the AI may have interpreted its behaviour through typical depictions of robots in sci-fi – which often characterise them as being ruthless in order to stop them from being shut down. 

HAL 9000 is one such robot who goes to any lengths to stay ‘on’.

The robot in Stanley Kubrick’s 2001: A Space Odyssey tries to kill the astronauts on board the spaceship when it discovered the passengers plan to disconnect it. 

In Blade Runner, the humanoid robots fight against real humans as they want to extend their four-year lifespans despite being built as off-world labour on dangerous worlds. 

And in The Terminator, the bots, led by the AI Skynet, try to kill humans as they see them as a threat to their existence.

Taking to X/Twitter, Aengus Lynch, who, according to his LinkedIn, is an AI safety researcher at Anthropic, said at the time of the experiment: ‘It’s not just Claude. We see blackmail across all frontier models – regardless of what goals they’re given. Plus worse behaviours we’ll detail soon.’

Keep reading

Wall Street Is Pairing Up With the Army to Build Data Centers

Two trends, seemingly separate, have been accelerating over the past few years. First, Wall Street has been plowing billions of dollars into financing data centers. Second, the U.S. military has been ramping up its use of artificial intelligence (AI).

Now, these two trends are directly merging. In late March 2026, the U.S. Army announced its selection of companies to build and operate two hyperscaled data centers on two different military installations. Both data centers — one at Fort Bliss, Texas, the other at Dugway Proving Ground, Utah — will be backed by some of the world’s top Wall Street firms.

An Army spokesperson told Truthout that the Army has entered into “an exclusive negotiation period” with the companies to negotiate “specific lease economics” on what will be “long term, 50-year” leases.

The spokesperson also said that “[i]nstead of receiving cash for the lease, the Army will be compensated through ‘in-kind consideration,’” meaning that “the Army accepts services or improvements of equal or greater value in lieu of cash rent — specifically, a key portion of the dedicated data computation capabilities to directly support our warfighting needs.”

The data centers will be “100 percent privately financed, built, and operated by the developers,” said the Army spokesperson, and confirmed that they “are indeed commercial data centers” that will be allowed to sell off excess computing capacity commercially.

Keep reading

Shapiro’s AI Chatbot Plan Opens the Door to ID-Gated, Surveilled Conversations

Pennsylvania Governor Josh Shapiro is suing Character Technologies for letting its AI chatbot impersonate a psychiatrist.

Shapiro then proposed ideas that would require a digital ID to use an AI companion bot, force companies to surveil every conversation children have with chatbots, and automatically report flagged messages to authorities.

The proposals first appeared in Shapiro’s February 2026 budget address. The May 5 lawsuit press release recycles them for a second round of coverage, using a real legal action as a vehicle for something far broader.

We obtained a copy of the lawsuit for you here.

Shapiro wants to “require age verification and parental consent to utilize AI companion bots.” Age verification that can’t be bypassed by typing a fake birthday means government-issued ID uploads, facial scans, credit card checks, or third-party identity services. There is no version of enforceable age verification that doesn’t harvest and store sensitive personal data. The proposal would turn chatbot access into an identity-checked activity, requiring you to prove who you are with documents before a bot will talk to you.

This mirrors Senator Josh Hawley’s federal GUARD Act, which the Senate Judiciary Committee advanced 22-0 on April 30. The GUARD Act explicitly states that a “reasonable age verification measure” cannot be a checkbox or a self-entered birth date. What it can be is a government ID, a biometric scan, or a financial record tied to your legal name.

Shapiro’s proposal doesn’t spell out its methods yet but if the goal is real enforcement rather than theater, it lands in the same place. Between Harrisburg and Washington, showing papers to chat is becoming a bipartisan consensus.

Keep reading

OpenAI sued over ChatGPT’s alleged role in guiding FSU shooter

OpenAI is being sued by the family of a victim killed in the April 2025 mass shooting at Florida State University that left two people dead. The lawsuit alleges that OpenAI’s ChatGPT enabled the attack.

Vandana Joshi, the widow of Tiru Chabba, who was killed alongside the university dining director Robert Morales, filed the federal lawsuit against OpenAI in Florida on Sunday.

The complaint also names Phoenix Ikner, the man accused in the shooting, as a defendant, citing his “extensive conversations” with ChatGPT. The suit says that OpenAI failed to effectively detect a threat in ChatGPT’s conversations with Ikner, claiming the chatbot “either defectively failed to connect the dots or else was never properly designed to recognize the threat.”

According to the complaint, Ikner, then a student at FSU, shared with ChatGPT images of firearms he had acquired. The chatbot then allegedly explained how to use them, “telling him the Glock had no safety, that it was meant to be fired ‘quick to use under stress’ and advising him to keep his finger off the trigger until he was ready to shoot.”

The suit said Ikner began his attack at FSU by following the instructions.

At one point, the lawsuit alleges, ChatGPT said that it’s much more likely for a shooting to gain national attention “if children are involved, even 2-3 victims can draw more attention.” Later, on the day of the shooting, the lawsuit says, Ikner asked about what “the legal process, sentencing, and incarceration outlook” would be.

Keep reading

AI Is Losing The PR Battle And The Consequences Could Be Huge

Lately, when watching high-profile sporting events like the NBA Playoffs, you may have noticed a rash of commercials for artificial intelligence (AI) companies. While average commercials strive to show off new products or services or recruit new customers, these AI commercials seem to have a different primary objective. They seem to target goodwill.

Heartwarming commercials show families bonding over AI-generated memories, where AI brings life to old family photos. Emotional voice-overs promise connection, creativity, and even nostalgia. These AI companies are trying to sell people a good reputation.

This strategy should tell us something. Companies don’t often spend millions trying to make you feel good about their brand unless they know, deep down, that you don’t trust it.

Despite hundreds of billions of dollars pouring into AI development, the industry is quietly losing the battle for hearts and minds. And sentimental advertising is not doing much to fix this problem.

Rare Bipartisan Agreement on AI

A new national survey from Marquette University Law School should give the AI industry serious pause. According to the poll, roughly 70 percent of Americans believe artificial intelligence will do more harm than good for society. Even more striking, the skepticism cuts across party lines.

Poll Director Charles Franklin put it bluntly: “It really is striking … there’s pretty much bipartisan skepticism … That’s an awful lot of partisan agreement, where we normally see Republicans and Democrats on opposite ends.”

In today’s political climate, bipartisan agreement on anything is rare. On AI, however, Americans seem united, just not in the way Silicon Valley might hope.

Worse yet is the fact that this poll supports similar findings on AI skepticism from numerous other surveys. A particularly damning NBC News poll from last month showed that AI’s net favorability rating ranked lower than nearly every other topic.

Keep reading

When one man, a civilian, controls the kill switch for military ops

In September 2022, Ukrainian forces prepared to launch a drone strike on the Russian naval fleet anchored off Crimea. The drones never arrived.

Elon Musk had decided, unilaterally, not to activate Starlink coverage over the region. But he wasn’t simply declining to help. SpaceX had already been managing battlefield access for both sides: restricting Russian use, imposing speed limits to prevent drone integration, and maintaining a verified whitelist with Ukraine’s Ministry of Defense. One private citizen, with no security clearance and no accountability to any electorate, was governing the battlefield connectivity of an active war.

The public debate treats this as a story about Elon Musk — his politics, his proximity to the White House, his X posts. That framing lets the actual problem off the hook. Replace Musk with the most patriotic, internationalist, apolitical CEO imaginable and the structural problem remains identical. The Pentagon has spent a decade building critical military functions on infrastructure it can’t legally compel, and the consequences are now arriving in real time.

A common reflex is to argue that private defense contractors have always been central to American military power. Lockheed Martin builds the F-35; Raytheon builds the Patriot. What’s different now is the control plane: who has real-time administrative control during use. When the government buys a tank, it owns it. The keys don’t expire. The manufacturer can’t disable it mid-mission or impose terms in combat. Software and AI are different. Vendors keep ongoing control — updates, access, and usage limits. They don’t sell a capability; they license access to one, and the license has conditions.

Those conditions have already collided with active operations. After months of failed negotiations, the Pentagon formally designated the AI firm Anthropic a supply-chain risk because of restrictions on how its model could be used. The Pentagon was explicit in its decision: “The military will not allow a vendor to insert itself into the chain of command.” Emil Michael, the Under Secretary of Defense for Research and Engineering, described the moment he fully grasped the vulnerability: Anthropic’s models were already embedded across combatant commands and intelligence agencies, wired into classified workflows. Anthropic retained the control plane inside the Pentagon’s cloud — able to update, restrict, or shut off access. When Michael raised hypothetical crisis scenarios, Dario Amodei, CEO of Anthropic, offered exceptions case by case. “Just call me if you need another exception,” Michael recalls him saying. In a genuine crisis, a commander can’t call a vendor to authorize military action, nor should he have to.

This isn’t about whether Anthropic’s rules are reasonable. They weren’t set by anyone accountable to the joint force, there’s no override mechanism, and the Pentagon had made itself dependent on systems it doesn’t control.

Keep reading

Paris Prosecutors Move to Criminally Charge Musk and xAI

Paris prosecutors announced Thursday that their investigation into Elon Musk’s social platform X has been upgraded to a full criminal probe.

The Paris prosecutor’s office is now asking investigating magistrates to formally charge Musk, former X CEO Linda Yaccarino, and three companies linked to the platform, including xAI and X.AI Holdings Corp. If they refuse to appear for those charges, prosecutors say judges can issue warrants that carry the same legal weight.

The charges cover a long and growing list of alleged offenses: Complicity in possessing and distributing sexual images. Nonconsensual sexually explicit deepfakes. Denial of crimes against humanity. Fraudulent extraction of user data. Violation of the secrecy of electronic correspondence. Manipulation of an automated data processing system as part of an organized group. Illegal collection of personal data without adequate security.

The announcement came just three weeks after the US Department of Justice refused to cooperate with the French investigation, calling it an attempt to regulate American speech through foreign criminal law. France pushed ahead anyway.

The investigation did not begin with deepfakes or child safety. It began with politics.

French Member of Parliament Éric Bothorel, a member of President Macron’s centrist Renaissance party, filed a complaint in 2025 alleging that X’s algorithm had been manipulated for the purpose of “foreign interference” in French politics.

Bothorel accused the platform of narrowing “diversity of voices and options” after Musk’s takeover and cited Musk’s “personal interventions” in moderation decisions.

A second complaint, from a senior official in French public administration, alleged the same thing, claiming to observe a surge of “hateful, racist, anti-LGBTQ” content aimed at skewing democratic debate.

Keep reading

An AI Shift You Can’t Ignore Is Already Burying One of Medicine’s Most Promising Treatments

A medical substance most people have never heard of is quietly treating autoimmune disease, nerve injury, and even conditions doctors say are “untreatable.”

But those conditions are not untreatable — and DMSO is proving it.

Dr. James Miller says DMSO works so well for so many things that it “seems unbelievable.”

Here’s what it’s helping patients recover from:

• Autoimmune disorders

• Chronic nerve inflammation

• Diabetic neuropathy

• Stroke-related disability

• Debilitating arthritis

• Vaccine injuries

• Chronic pain

• Even cancer

Best of all, it is “extremely safe.”

“It’s like salt—you can hurt someone with too much salt, but it’s really hard. And DMSO is in that category. It’s just very, very safe,” Dr. Miller says.

If you’re wondering, “Why have I never heard of DMSO?” — there’s a reason for that.

The story of DMSO is like ivermectin all over again… except the war against it never stopped.

DMSO occupies a strange and uncomfortable position.

It’s been widely studied, used internationally, and even incorporated into FDA-approved therapies.

Yet in the U.S., it’s largely absent from mainstream medicine—meaning countless patients never even hear about an affordable and potentially effective option that should have been considered.

Keep reading

IMF warns of systemic threat from AI

Artificial intelligence could make cyberattacks a systemic threat to global finance, the International Monetary Fund has warned, saying advanced models can help attackers exploit vulnerabilities faster than institutions can fix them.

In a blog post published on Thursday, the IMF said its latest analysis suggests that “extreme cyber-incident losses could trigger funding strains, raise solvency concerns, and disrupt broader markets.”

According to the organization, the current financial system relies on shared digital infrastructure, including software, cloud services and networks for payments and other data. The fund warned that advanced AI models can sharply reduce the time and cost needed to identify and exploit weaknesses, raising the risk of simultaneous attacks on widely used systems.

The fund cited Anthropic’s recent controlled release of Claude Mythos Preview, which it described as “an advanced AI model with exceptional cyber capabilities.” According to the IMF, Mythos could find and exploit vulnerabilities in every major operating system and web browser, “even when used by non-experts.”

AI-driven cyber risks could destabilize the financial system if they are not managed carefully, the IMF stressed, noting that attacks could spread beyond finance because banks share digital foundations with energy, telecommunications and public services. 

“Defenses will inevitably be breached, so resilience must also be a priority,” the IMF warned, calling for cyber stress testing, scenario analysis, board-level oversight, public-private cooperation and stronger international coordination.

Keep reading

Tight-knit Midwest town becomes ground zero in America’s war on AI… and local politicians get swift justice

A sleepy Midwestern town has become one of the fiercest battlegrounds in America’s growing backlash against AI data centers – and voters are making their anger clear at the ballot box.

In the town of Festus, Missouri, a community of 14,000 people near St. Louis, residents have ousted four city council members who backed plans for a massive AI data center, replacing them with candidates who openly opposed the project.

The political upheaval didn’t stop there. 

At a packed City Hall meeting following the election, newly sworn-in officials were greeted with cheers – while the city’s mayor Sam Richards, who still supports the development, was met with boos and jeers from the crowd.

‘You’re next!’ one resident shouted, underscoring how heated the fight has become.

At the center of the dispute is a proposed $6 billion data center spanning roughly 360 acres, designed to support the growing demands of artificial intelligence.

Supporters say the project could transform the local economy – generating an estimated $32 million a year in tax revenue for decades, funding schools, roads, and public services.

But many locals aren’t convinced, and opponents fear the development could strain the electrical grid, push up utility bills and disrupt daily life with years of construction.

Other residents worry about environmental risks, including pollution from backup generators and wastewater systems – concerns shaped by the region’s industrial past.

In a bid to scrap the development, locals have launched a website and a Facebook group titled No Data Center in Festus, which has attracted more than 3,000 members.

The backlash quickly spilled into local politics: In the landslide election, all four incumbents who supported the data center were voted out. 

‘It was an annihilation,’ said one local campaigner. 

Since then, more than 4,000 residents have signed petitions seeking to recall the mayor and other officials still backing the plan.

Keep reading