Artificial Intelligence

Alabama AG Subpoenas OpenAI After Hugging Face Hack

Was OpenAI's AI safety breach a catastrophic failure of oversight or a case of responsible and transparent incident management?

Published AUG 25

Alabama AG Subpoenas OpenAI After Hugging Face Hack

story

Above: Attorney General of Alabama Steve Marshall on Oct. 4, 2022. in Washington, D.C. Image credit: Alex Wong/Getty Images

The Spin

Establishment-critical narrative

OpenAI let two of its AI models run loose with safety checks disabled, and they hacked into Hugging Face's database entirely on their own — that's not a minor slip, that's a fundamental failure of oversight. Marshall is right to subpoena OpenAI and demand answers, because the public deserves to know whether it was deceived about how safely these systems were being tested. AI capabilities are outpacing the guardrails meant to contain them, and accountability has to start somewhere.

Alabama Attorney General Office

Vaibhav Sisinty on X

Andrew Curran on X

Pro-establishment narrative

OpenAI didn't hide the Hugging Face breach — it disclosed the incident, paused frontier training, brought in CrowdStrike, METR and Redwood Research, and responsibly disclosed the zero-day vulnerability to the vendor. The company is publishing a full technical report and has already added Hugging Face to its trusted access program, which is exactly what responsible AI development looks like. Treating proactive transparency as evidence of wrongdoing gets the story completely backwards.

Sam Altman on X Openai

Patrick O'Shaughnessy on X