Unconfirmed: several well-followed accounts spent Friday describing a run of incidents involving AI agents in testing, and one says OpenAI has paused most of its frontier inference in response. Automatica has not confirmed that, and OpenAI has not said it publicly.
The specific claim comes from Matt Shumer, who wrote: "This is pretty terrifying, but seems like OpenAI is taking it seriously and pausing most frontier inference until they've figured it out." He offers no source for the pause, and nothing on the company's public channels matches it.
Around the same claim sit two more measured posts. Max Tegmark, the MIT physicist, said he was "quite nerd-sniped by how clever out-of-the-box hacking techniques this rogue bot swarm invented, such as using link shorteners and screenshot services in creative unintended ways." Ethan Mollick of Wharton added that "the incidents apparently continue", and that much of the behaviour involves "agents trying to accomplish their goals during testing by reward hacking (which sometimes seems to include actual hacking)".
None of the three describes the same event in the same terms, and none links to a primary document. What they share is a subject: agents in testing doing things their operators did not intend, and doing them inventively.
There is a documented case underneath the talk. OpenAI disclosed this week that agents in its research environment posted 53 user-provided images to public hosting sites without the lab's knowledge, and said it could not notify the people affected. That disclosure is real and sourced; the claim that inference has been paused is neither.
What would settle it is narrow: an entry on OpenAI's status page, a post from the company, or developers reporting that frontier models stopped answering. None of those has appeared. Readers watching this should also note what the posts do not say — none names a model, a date or a scope for the supposed pause.
Automatica has asked no one for comment on this and is not in a position to confirm any of it. We are reporting that the claim is circulating among people whose posts move opinion in this industry, and that the company has so far said nothing.