Post

Simon Willison
@simonwillison.net

I wrote about the completely wild incident where OpenAI were testing a new model and it broke out of its sandbox and broke INTO Hugging Face to steal the answers to the benchmark simonwillison.net/2026/Jul/22/...

simonwillison.netOpenAI’s accidental cyberattack against Hugging Face is science fiction that happenedThis story is wild. The short version: OpenAI were running a cybersecurity test against an unreleased model, with the model’s guardrail features turned off. Rather than solve the test, the …
Simon Willison
@simonwillison.net

Tucked away in this article is an appeal to the AI skeptics to PLEASE stop writing off stories like this OpenAI accidental exploit of Hugging Face as a dishonest marketing trick Frontier models can find and exploit vulnerabilities now, it helps nobody to pretend that they can't!

Sam Rose
@samwho.dev

If I had built a system capable of autonomously breaking out of a sandbox and hacking another company’s systems, you can bet your arse I’d use it as marketing. It’s the most awesome marketing opportunity in the world. Scary, but awesome. I’ve no doubt it happened, so it’s not dishonest.

0 reposts10 likes

Luke
@lukeocodes.dev

If I had open source competitors now catching up of over taking me just before I IPO I’d be bullshitting about having something special too. In every direction, Anthropic and OpenAI are being caught with their pants down because they’ve hit an inflection point of scale, resources, and cash flow

cephyn
@cephyn.bsky.social

But in OpenAI and HF's discussion, they explicitly pointed out that HF detected and countered the breach using open weight models - because the closed models refused to engage due to security guardrails. So...that argument doesn't make a lot of sense.

Luke
@lukeocodes.dev

It makes total sense. If they drum up enough hype about security risk of frontier models, the orange turnip and his mates will do something silly like ban their use entirely in the US unless under contract by the GOV. OpenAI and Anthropic already have contracts so they’ll be fine

Luke
@lukeocodes.dev

It also isn’t lost on me that both Anthropic and OpenAI are hinting here that their models are edging towards AGI which is just fucking nonsense. J-Space and this breach? For starters why aren’t they’re airgapping their sandbox, and if they did then I want to laugh even harder at them

Benjamin
@premium-content.bsky.social

There must be ways of making the point without presenting your company as an uncalculable liability risk. Making the best of screw-up is one thing. The conspiracy theories about this being an intentional stunt don't add up.