As noted above, we have deployed the classifier as an experimental addition to our Safeguards framework, monitoring a percentage of Claude traffic. Its real-world performance has confirmed that the classifier works effectively beyond our testing environment. Whereas our synthetic test data provided clear examples of harmful and benign exchanges, the distribution of actual user traffic proved more complex and surprising, yet the classifier still performed well. One example of how real-world deployment differs from testing is that the classifier flagged certain conversations about nuclear weapon
Developing Nuclear Safeguards for AIOSes Nanny state discovers Linux, demands it check kids' IDs before booting Age-verification laws target operating systems because apparently teenagers having root access is now a safeguarding crisis OPINION A new…
…A new app just released on the App Store is the perfect example of safeguards should be implemented when Apple launches its smart glasses. Meta Ray-Ban Display. Image source: Meta The…
…Opportunities for engagement on AI safety Anthropic supports international AI safety dialogue with AI experts in China, when possible. The world has a vested interest in safe AI, regardless of where it…
Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
Everything announced at Apple's WWDC 2026 keynote Siri AI is finally almost here and Liquid Glass will look a little less liquid soon. By Kris Holt Updated: June 8, 2026 3…
…the risks—and benefits—of AI models for cybersecurity. We stated that we would keep Claude Mythos Preview’s release limited and test new cyber safeguards on less capable models first. Opus…
Vibe coding is great for the App Store economy, but Apple is still wary about its use without safeguards in place. It's a fine balance that's going to…
…It is calculated based on the actor's activity across Claude.ai, Claude Code, and our API, drawing on our safety classifiers alongside open-source and internal threat-intelligence indicators. The higher…
…Training models to recognize and act on their own uncertainty is an important safety property that complements external safeguards like permission systems and human oversight. At Anthropic, we train Claude to ask…
…May 25, 2026, Pope Leo XIV released an encyclical on the topic of AI: "Magnifica humanitas: On safeguarding the human person in the time of artificial Intelligence." Anthropic co-founder Chris Olah…