API Vulnerabilities Expose Hidden Reasoning Tokens Across Frontier AI Models

davidmanheim · x · 2026-08-11

Security researchers have discovered a vulnerability in the APIs of major frontier AI companies that allows for the extraction of hidden reasoning tokens. They verified that the extracted reasoning token count matches the billed API thinking tokens 1:1 for most queried prompts.

Experts criticized the abysmal security mindset of frontier AI companies, noting that this vulnerability directly threatens their most valuable intellectual property, beyond just standard safety concerns.

Related event: Study Reveals API Flaw to Extract Encrypted Reasoning Traces and Evidence of Distillation(38 posts)→

Original post →

More from Safety

Safety channel →