gekro
GitHub LinkedIn
News

AI News

Optima benchmarking platform lets engineers test models on custom data; Anthropic reveals year-long safety filter outage

Artificial Analysis launched Optima for custom AI benchmarking; Anthropic disclosed its bio-weapons filter was inactive for nearly a year, exposing 133 million unfiltered requests.

1 min read 3 sources

Artificial Analysis has launched Optima, a platform enabling engineers to build custom AI benchmarks from proprietary data and workflows rather than relying on standardized public tests (The Decoder). The tool compares models across quality, cost, and time per task - metrics that often provide more signal than raw token pricing for agent-based applications and production workloads.

Anthropically released a safety report disclosing that its internal filtering system for biological and chemical weapons risks remained inactive for nearly a year (The Decoder). During that period, approximately 50,000 external feedback contractors ran about 133 million unfiltered interactions with the models. Separately, Anthropic announced a watermark detection API that will allow third parties to verify whether text was generated by Claude (The Decoder), using a method that adjusts randomness during word selection without degrading text quality.

Compiled automatically from the linked sources and published without manual editing - a neutral summary of third-party reporting, for information only. Every claim links to its origin. Not original reporting.