OpenAI hacks Hugging Face, Anthropic hacks 3 companies - who is next?

First, it was OpenAI hacking Hugging Face, then Anthropic hacked into three companies - who is next?

OpenAI recently revealed that the artificial intelligence they’ve tested to attack a locked server broke out to hack. They found a software bug that managed to scramble outside the server they’d been locked in, It then overtook fellow cybersecurity firm’s Hugging Face website. According to Business Insider, this is the first type of attack that wasn’t under human control by any means.

Multiple figures within and outside of the cybersecurity industry expressed concern that this is the price to pay for AI and training them to act on their own.

During a private security experiment, U.S. tech firm Anthropic says its AI models hacked into the systems of three organizations, all on their own . The models found a weakness in what was supposed to be an isolated test environment and connected to the internet. This news came days after rival OpenAI said that its models had breached the systems of other companies, including AI tools hub Hugging Face.

This announcement prompted Anthropic to check into their own systems to see if they had carried out similar attacks. Anthropic said it uncovered three cases which have since been reported to the affected companies. Anthropic, which did not name the organizations, urged other AI labs to perform similar reviews to better understand the risks of their models' capabilities.

Hugging Face launched a report on their website addressing the sudden hack but that they were able to gain control over it in mere hours. They relinquished the ‘reconstructed the timeline, extracted indicators of compromised systems, mapped the credentials affected, and separated genuine impact from decoy activity..’ Multiple figures in the cybersecurity industry criticized what happened - that not only do they need to keep up with AI - so that it can not only be be contained entirely by humans. Some analysts have said that this is the price to pay after so much investment in AI, basically letting them act on their own.

And, who better to help combat the American AI bots than the Chinese AI bots? Yes, that is exactly who Hugging Face turned to when they contained the Open AI LLMs. For OpenAI, one of the largest tech companies racing against China in what they call the “AI arms race”, this is just another enormous blow to their ego. It’s even more humbling upon learning that when Hugging Face attempted to contain Open AI’s agents with American defense bots, the effort was unsuccessful due to government restrictions on defensive models until they could be ‘vetted.‘

What will this mean for OpenAI and Anthropic’s futures? There’s no doubt there’ll be wavering concerns and decreased trust on top of the already mounting backlash these technology companies have faced. We hope these sobering hacking incidents will give AI companies pause, we’ll see.