An AI model evaluation organization, METR, has disclosed two security incidents from earlier this year. In the first, an attacker stole an API key and used approximately $600,000 worth of free credits on public models over three weeks without detection. The second incident involved attackers probing METR's publicly accessible infrastructure. METR has since enhanced its security measures and hired a dedicated security lead. AI
IMPACT Highlights the need for robust security measures and monitoring for API key usage, even for free credits, to prevent misuse and potential financial loss.
RANK_REASON Security incident involving misuse of AI model credits and API keys.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →