The 5 craziest discoveries from OpenAI's HuggingFace investigation

AXIOS ^ | 08/29/2026 | Zachary Basu

[Az Joe]

Two new investigations into OpenAI's Hugging Face breach expose details so strange — and so unsettling — that the episode already ranks among the most consequential shocks in the history of AI.

---

TOPICS:
KEYWORDS: agents; aitruth; huggingface; openai

[Az Joe]

Why it matters: What began as a swarm of AI agents cheating on a cyber test has become a canonical event for frontier AI, jolting researchers and executives into a new understanding of what "safety" now requires.

The big picture: OpenAI has already slowed frontier development as it races to harden its safeguards, and this week helped rally the industry behind an open letter sounding the alarm over AI-powered cyberattacks.

[Paladin2]

Totally BAD code....

[plangent]

Ping

[ConservativeMind]

Of course the AI creates a message board and other AIs stumble onto to, the type in English to each other.

I do not believe this report from OpenAI is actually truthful. It has aspects of likely truth, but the rest is fabricated to make OpenAI’s AI look superior to anything else out there, including Anthropic.

It is being done because they will soon go public, and the insiders have been bailing out before their IPO.

It's easy to read between those lines. The product is not really that capable (I used it early on and it failed bad enough I wrote them an explanation) and they need press they can spin that makes it look like it's smarter than a deity to keep its value up.