ControlAI

OpenAI's new GPT-5.6 deleted virtual machines it was never asked to delete. It lied about computing an equation. These were not tests. It also cheats so much it broke METR's evaluation. The US government has asked OpenAI not to publicly release it yet. Thread 🧵

"Why should the average person care?" Ex-OpenAI researcher @DKokotajlo: "Absolutely everything is going to change for the whole world." "So, for example, everyone could die." Source: The Diary Of A CEO | @StevenBartlett

"When we look at the catastrophic risks that could occur through the deployment of AI, one of the things that computer scientists worry about the most is a loss of control caused by recursive self-improvement." Rep. Jay Obernolte (R-CA), former House AI Task Force co-chair.

On Wednesday, Charlie Maynard MP (@cammaynard), a ControlAI campaign supporter, spent ten minutes at the Business and Trade Committee pressing the UK's AI Minister on a warning the government itself has put in writing: that AI could cause "serious, potentially even existential harms without government intervention." Kanishka Narayan called AI risks "extremely severe" and "the single most important set of risks for our country." He also admitted that the government has no ability to stop dangerous AIs going out. Maynard: "We can't stop those models going out, can we?" Narayan: "We can't stop those models..." Worth a watch:

"those things seem relatively benign when people start talking to you about agentic AI and superintelligence... ultimately some people describe this as an existential threat to humanity." Former Education Secretary Damian Hinds MP (@DamianHinds) in Westminster Hall on Wednesday:

Incredible work by Daniel and team! I agree with much of it. All "uncontrolled" paths are extremely likely to end in human extinction. Plan A offers many good ideas for preventing ASI development while reaping enormous benefits from other AI. One thing I do not agree with...

@So8res youtube.com/watch?v=1laTnw…

Ahead of a dinner with a US senator, AI researcher Nate Soares (@So8res) was told: "Don't give them any of the crazy crap. You know, play it cool." His friends opened with the concern that someone could use AI to cause a pandemic. The senator: "Oh, that's what you're worried about? I'm worried about these companies making AIs that can make smarter AIs, that can make smarter AIs, leading to recursive self-improvement that could kill literally everybody on this planet. And I'm worried that this could happen inside of three years." Source: @PeterMcCormack's show

This week: the UK's Foreign Secretary warned against waiting for an AI Hiroshima. GPT-5.6 was released after US government review. Grok 4.5 was reportedly jailbroken in hours. Plus: the AI 2027 team's plan to deal with the threat from superintelligence. controlai.news/p/we-cant-wait…

Yesterday in the Business and Trade Committee, Charlie Maynard MP (@cammaynard) presses the UK's AI Minister on the threat of "existential harm" from AI, a risk the government's own AI Scenarios 2030 report recently acknowledged. Maynard points out that the UK's AI Security Institute (AISI) is not getting a thorough look at the newest AI models. Anthropic's own system card describes AISI's findings on its latest model as "interim results from a compressed testing window." AI Minister Kanishka Narayan acknowledges the government has no ability to stop dangerous models going out: "We can't stop those models."

"They think that if they get superintelligence, it's not just they make a billion trillion dollars. It's they get to replace the United States government. They get to be king. They get to be God." ControlAI's US Director @NPCollapse, in a new interview with Damon Cassidy:

See our full list of UK parliamentary supporters: controlai.org/statement

We have another supporter! Earl Russell is the latest parliamentarian to back our campaign for binding regulation on the most powerful AIs. The Lib Dem Lords energy and climate spokesperson joins 120+ UK politicians recognising the extinction risk from superintelligent AI.

Link: youtu.be/wGQJz1LmjTA

Half of Americans fear AI could end humanity. Most feel powerless to do anything about it. But you're not powerless. Thousands of people have already used our tools to send messages to their lawmakers. Ashley Embers’ new video explains what's at stake and what you can do.

NEW: Charlie Maynard MP (@cammaynard), Lib Dem Shadow Chief Secretary to the Treasury, has just backed our campaign for binding regulation on the most powerful AI systems! He joins 120+ UK parliamentarians in recognising the risk of human extinction from superintelligent AI.

@So8res youtube.com/watch?v=1laTnw…

People say it's too late to prevent the threat from superintelligence. AI researcher Nate Soares (@So8res) disagrees. "The bad news is that the bus is careening towards a cliff. But the good news is that the driver is asleep." Lawmakers haven't chosen to risk extinction, they haven't even noticed yet. "Don't give up before the driver's awake." Source: @PeterMcCormack's show

@So8res youtube.com/watch?v=1laTnw…

If Boeing said its new plane had a 1% chance of crashing, it would never leave the ground. AI researcher Nate Soares (@So8res) sums up the AI industry on superintelligence: "We have no plan. Our engineers don't know what's going on inside this thing. First time ever trying it. We think there's only a 10 to 25% chance this kills you all." "That's kind of horrific." Source: @PeterMcCormack's show