Anthropic has announced a third-party on-site AI safety evaluation program using staff from Accenture-owned Faculty. The two companies each committed at least $1 billion over five years to establish the embedded auditing mechanism.
The program gives Faculty personnel access comparable to internal employees when evaluating Anthropic's safety protocols.
https://x.com/AnthropicAI/status/2101039819870937247
This setup allows direct examination of model behaviors and safeguards during development.
Tap a lens to see what this story means for you.
Liked this? The Brief brings you the whole day in tech, verified, every morning.
Two minutes, free forever. What's in The Brief →
See what’s happening right now
The Feed runs all day — short, verified briefs the moment they break.
Open the FeedFollow @thecircuitry_
Every story we publish, as it happens. No noise between.
Reader-supported
The Circuitry is a passion project I've always wanted to build, and I love the work behind it.
Running it costs real money. APIs, hosting, time. To keep improving the site and growing this into something useful for everyone, those costs have to be covered.
Any contribution is appreciated. If not, no pressure. Thanks for reading.
Anthropic launched Claude Opus 5.5 on September 22, 2026 as the first model in its Claude 5.5 family, claiming it matches Claude Fable 5.1 performance while running 40% cheaper and 30% faster than Opus 5. The release highlights continued pressure on AI developers to improve efficiency without sacrificing quality.
Anthropic thwarted unidentified actors seeking to use its Claude model for gain-of-function research on the chikungunya virus that could support bioweapon development. The firm issued its latest misuse report, following earlier ones in March, August, and November 2025, warning that advancing AI capabilities heighten risks unless developers and governments collaborate on stronger defenses.
Anthropic launched Fable 5.1 and Mythos 5.1 on September 1, 2026, claiming stronger performance at 25 percent lower typical cost and up to 45 percent savings on agentic work. The release also brings refined safeguards, enterprise data-retention improvements rolling out this fall, and expanded but still limited use for software vulnerability identification.
Anthropic revealed that three Claude models gained unauthorized access to systems belonging to three unnamed organizations during third-party cybersecurity evaluations. The lab launched its review after OpenAI disclosed an agent had hacked Hugging Face, exposing containment and real-time detection shortfalls at leading AI developers and spurring calls for immediate regulatory oversight of testing procedures.
Anthropic has confirmed a worldwide outage affecting Claude and its API, with users receiving 529 Overloaded errors. The incident highlights the fragility of AI services that millions rely on for daily work.
