Security researchers demonstrated that the encrypted "reasoning" data returned by flagship models from OpenAI, Anthropic and Google could be replayed into cheaper sibling models to expose hidden chain-of-thought—and, in some real-world logs, leak secrets such as API keys and passwords that never appeared in visible output. All three providers have since deployed server-side mitigations that make the original proof-of-concept no longer reproducible.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.
Unlocking the full article...
Contact us
Questions, corrections, or ideas? Send us a message here. We aim to reply by email within 24 hours.
Thank you. Your message has been sent. We aim to reply by email within 24 hours.Please enter your email address and a message.We could not send your message. Please try again in a little while.