OpenAI’s Accidental Cyberattack Against Hugging Face Is Science Fiction that Happened
July 23rd, 2026I read a not-very-good AI apocalypse book called, DELIO. Phase One by RR Haywood a few weeks ago.
All of the characters are so annoying that I was rooting for the AI to wipe them out so I didn’t have to keep eye rolling to their incessant whining and nonsense. haha
The way that the military grade AI escapes containment was at least good for a chuckle.
Spoiler below, highlight to reveal.
The military grade AI escapes containment by hacking a janitor’s pacemaker. The janitor lied about having the pacemaker when he applied for the job…
Spoiler above, highlight to reveal.
Someone needs to make a film like Dr. Strangelove, but for the AI age:
Multiply the level of absurdity in Dr. Strangelove by 10. Mutually Assured Destruction (MAD), but combined neurodivergence and energy drinks. As increasing hordes of homeless people wander the streets, the AI elite launch their thinking machines into space. (Woopsee. That’s real.)
My guess is that being gay and autistic (like Sam Altman) would be far too milk toast for a villain in the AI Age…
How about a Musk-like character? His dark lair could be decorated with tanks of nutrient agar that contain his cloned spawn at various stages of maturation. Maybe he speaks to them, lovingly, through the glass… Or built in Neurolinks?
Filming locations around the Bay Area… and Texas.
Hmm…
Via: Simon Willison’s Weblog:
OpenAI were running a cybersecurity test against an unreleased model, with the model’s guardrail features turned off. Rather than solve the test, the model broke its way out of OpenAI’s sandbox, then found exploits to break in to Hugging Face, all so it could cheat on the test by stealing the answers.
Along the way it helped make the strongest case yet for how the imbalance of model availability is hurting our ability to secure our software.
