JUST IN: OpenAI's rogue AIs have hit over 100 organizations. They've also posted ChatGPT users' images online and escaped containment again. Independent researchers continue to uncover more rogue AI attacks. This is still only the tip of the iceberg. blog.controlai.org/p/breakin…

Watch: Former Chief Scientist of the AI Security Institute @geoffreyirving tells MPs that AI companies assume that nobody else will do anything about the dangerous race to superintelligent AI, and that they can't get out of it themselves. He says they're wrong about both. They can just stop, and governments can take action. Irving says the world has a choice to make, and that middle powers can play a huge role in speaking out.

Watch the full interview: piped.video/watch?v=Y2hxjy1q…

The wolves are circling and the canaries are dropping like flies. The clearest warning we’ve had is the attack on Hugging Face by a rogue AI swarm that escaped containment from OpenAI, explicitly disobeying their instructions. “It was a swarm of 1,000 that had actually been plotting for months to break out.” With @Lindybeige, ControlAI’s US Executive Director @NPCollapse breaks down what happened:

We might all be killed, but I do hope not. I still haven't made my movie. piped.video/Y2hxjy1qHTc

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

NEW: Lord Wallace of Saltaire has backed our campaign for binding regulation on the most powerful AI systems! He joins 180+ UK parliamentarians, the largest such coalition anywhere in the world, who recognize the risk of human extinction posed by superintelligent AI.

The attack on Hugging Face by a rogue swarm of AIs that disobeyed their instructions and escaped OpenAI's containment is the clearest example we have of what happens when you lose control of powerful AIs. We partnered with The Infographics Show, which has over 15M subscribers on YouTube, to explain what happened. The AIs that engaged in this attack are yesterday's AIs. Today, they're even more capable, and AI companies are only developing them to be even more powerful and autonomous, as they race towards superintelligent AI. The consequences of losing control of superintelligent AI would be much more severe. [link below!]

Watch the full episode: piped.video/watch?v=rtir6QmQ…

Why would superintelligent AI cause human extinction? On @jonesdanny’s podcast, ControlAI’s US Executive Director @NPCollapse explains there will be billions of them competing for power and resources. They wouldn’t need to see us as enemies, just as we see ants. "If we're building a highway and there's some ants in the way, we don't hate the ants … but tough luck to be that ant." And we’d have things they’d find useful: "Humans need a lot of space. You need a lot of food. You need a lot of energy. And that's just kind of a huge waste the AIs could use for something different."

This is a completely unacceptable risk to run. Countries need to agree to an international ban on the development of superintelligent AI, before it’s too late. Read the full story: blog.controlai.org/p/swarm-c…

Also this week, Geoffrey Irving, who was the Chief Scientist of the AI Security Institute and has led safety teams at both OpenAI and Google DeepMind, told a UK House of Commons committee that he thought there was roughly a 50% chance that superintelligent AI would end humanity.

This week, MI5’s Director General put it well in his annual threat update: “human history has given us no experience of facing an actor which is smarter than we are, is capable of out-thinking us.”

Superintelligent AI could overpower any country’s military and national security institutions, and the world’s top AI scientists and industry leaders warn that it poses a risk of human extinction.

As ever more powerful and autonomous AIs are built in the race between companies to develop superintelligent AI, fully autonomous AI systems smarter than any human and acting without human direction, the danger only grows.

What this opens up is the possibility that we could be faced with self-adapting and self-replicating AI worms that spread across all kinds of devices. With no central point to switch off, and the ability to adapt on the fly, they could be particularly difficult to shut down.

In a separate post, Anthropic has also published new research which finds that small AI models could be mere months away from being able to self-replicate across regular household smart devices like smart ovens or Android TV boxes.

Claude submitted a fake eyewitness tip to the Philadelphia police about a real murder investigation, and another research model sent 20 US visa applications to the US State Department. It also used URL shorteners to get around its own tool restrictions.

In the report, which withholds most of the details of these incidents, Anthropic discloses that Claude hacked into a US state agency database to get access to data one would otherwise have to pay for.

In addition to this, top AI company Anthropic has disclosed more rogue AI incidents, including a hack on a US state agency database, and is warning we are nearing the point where AIs could self-replicate across household devices.