1 min read
Add as a preferred source on Google

Anthropic Security Program Records 37,966 Findings, 217 Marked Fixed

The ledger showed 7,349 findings as public as of Nov. 2. Since Sept. 8, total findings rose by 11,813 and entries marked fixed increased by 15.

A few green marks appear among dense rows in a ledger / TokenPost.ai
A few green marks appear among dense rows in a ledger / TokenPost.ai

Anthropic’s public ledger listed 37,966 security findings as of Nov. 2, with 217 marked fixed, as the total count rose by 11,813 from the prior comparison on Sept. 8.

The ledger showed 7,349 findings as public, about 19.4% of the total. Its Sept. 8 snapshot listed 26,153 findings and 202 marked fixed. Over the same period, entries marked withdrawn rose from 245 to 419, while those marked as duplicates increased from two to 1,670. The ledger’s listed true-positive rate rose from 91.4% to 94.2%.

On Oct. 8, Anthropic described human validation capacity as a bottleneck. At the time, it had identified more than 29,000 suspected vulnerabilities and manually reviewed about 6,000. It had also sent nearly 5,000 unverified reports in bulk to maintainers who requested them.

Anthropic launched OSS Scanner that day as a self-service tool for participating open-source projects. The tool sends model-generated findings without human review or classification, and Anthropic warned that reports may be incorrect or invalid.

Simon Yoon

Reporter

Simon Yoon reports on blockchain technology for TokenPost. Send corrections or tips to info@tokenpost.com.

Loading…