Anthropic, a US-based AI company, expanded its Cyber Verification Program for Claude into three tiers by combining it with Project Glasswing.

The program gives vetted security professionals broader access with fewer cyber-related blocks on the Claude Platform, Google Cloud Vertex AI, and Microsoft Foundry.

Defense Access covers incident response, malware reverse engineering, and vulnerability analysis.

Red Team Access also allows authorized penetration testing, but Anthropic said it would still block actions that could cause physical harm or mass disruption.

Specialized Access is for verified organizations authorized to test sensitive safety systems, with reviews conducted alongside the US government.

Anthropic generally requires data retention to monitor misuse.

An exception applies to organizations already using Claude Fable 5.1 or Claude Mythos 5.1 with zero data retention.

Enterprise Frontier Safeguards, which Anthropic said would combine zero data retention with cyber safeguards, is planned for later this fall.

OpenAI has introduced trusted-access channels for cyber defense and runs a researcher program.

The Frontier Model Forum has outlined frontier-model cyber safeguards, while the New York State Department of Financial Services has warned of heightened cybersecurity risks.

🔗 Source: Anthropic

Recent Anthropic developments