OpenAI, Anthropic, Google API Flaw Let Weaker AI Models Decode Stronger Models' Reasoning

A newly disclosed flaw in the way OpenAI, Anthropic, and Google carried hidden AI reasoning between API calls let researchers recover internal reasoning and secrets from session logs, including API keys and passwords. The weakness affected encrypted reasoning objects used by the providers' reasoning APIs, where a block created in one session could be replayed into another and, during testing,

The Hacker News

1d ago

1 0

Discussion

Be the first to comment

Add your perspective to get the discussion started.

No comments yet.

Be the first to share your take and keep the conversation moving.

Join the conversation

UPVOTERS

Community appreciation

See who found this content valuable and showed their support.

Lars Khan

TOPICS

Explore the same topics

Discover more content from the topics this post is mapped to.

bleepingcomputer.com

White House taps security firms for offensive hack-back operati…

A new White House memo signed by U. S. President Donald Trump instructs the National Coordination Center (NCC) to establish a program that would allow private…

Lars Khan

8h ago

thehackernews.com

Attackers Exploit SharePoint Authentication Bypass After Public…

Threat actors have begun to exploit a newly disclosed Microsoft SharePoint vulnerability following the release of a proof-of-concept (PoC) code. The vulnerabi…

maya af

9h ago

darkreading.com

'Jewelbug' APT Balances State Espionage & Cryptocurrency Th…

Researchers discovered hackers-for-hire performing cyber espionage and financially motivated heists from the same Web panel.

Dark Reading

11h ago

bleepingcomputer.com

"City-Forum" data-theft attacks target Salesforce, ServiceNow p…

An ongoing data theft campaign uses custom tools to steal data exposed to anonymous users through Salesforce Experience Cloud and ServiceNow customer portals.…

Nirano

15h ago

darkreading.com

Long-running Data Theft Campaign Targeting Salesforce, ServiceN…

The "City-Forum" campaign has been active since at least March 2025 and has targeted organizations across multiple sectors with custom tooling.

Dark Reading

21h ago

thehackernews.com

Attackers Exploit VMware vCenter Vulnerability to Gain Persiste…

Threat actors have begun to actively exploit a recently patched critical security flaw in Broadcom VMware vCenter, according to new findings from QUIRSO. The…

The Hacker News

1d ago

Keep browsing

Explore more from this topic

Dive into the full feed of curated posts covering Cybersecurity & Data Protection.

Browse Topics