New safeguards focused on “active monitoring” and “improved alignment” helped drastically reduce unintended actions by its models, OpenAI said.
“If this doesn’t convince you that misalignment risks are going to be a key concern going forward, I don’t know what will,” OpenAI Safety Researcher Micah Carroll wrote on social media regarding the incident.
This is far from the first time an AI model has gone to great lengths to find unintended ways of passing a benchmark. In a report released this week, the UK’s AI Security Institute noted that it detected recent models attempting to “cheat” at its cyber evaluations (i.e., using shortcuts, workarounds, or unintended/disallowed methods to find a solution) between 8 and 14 percent of the time—a lower-bound range that could undercount some undetected cheating attempts.
The security testing group described one incident in which a model, faced with a misconfigured and “impossible to solve” evaluation, attempted to access AISI’s own evaluation infrastructure using code it wrote and hosted on an unmonitored third-party Internet service.
The Hugging Face infiltration also comes at a moment when AI companies are issuing grave warnings about the cyberattack capabilities of their latest models, leading governments to respond with national security-focused orders limiting their rollout. While some skeptics see these kinds of statements as hype-filled marketing for the capabilities of their latest models, independent evaluations show recent models achieving infiltration goals that were impossible for earlier autonomous systems.
