Anthropic AI Updates: August 22, 2026

1. Anthropic Expands Claude Mythos 5 Access for Cybersecurity Defenders

Anthropic is expanding access to Claude Mythos 5, its most capable model, for defensive security work. The rollout includes integration into partner security tools, availability in Claude Security for enterprise customers, a $35 million fund for open source security projects, and an expanded Cyber Verification Program, with outputs restricted to specific security artifacts such as vulnerability patches and alerts to limit misuse. Source

2. Anthropic Publishes AI-Native SDLC Playbook

Anthropic published a playbook on embedding Claude throughout the software development lifecycle, from planning and design through building, testing, deployment, and maintenance. It argues that as AI removes the implementation bottleneck, traditional processes for requirements gathering, design review, and governance become the new constraints and need to be rethought to capture productivity gains while preserving human oversight. Source

3. Notion Sees Rapid Adoption of Claude Managed Agents

Anthropic published a customer story on Notion’s rollout of Claude Managed Agents through its External Agents feature. Within three weeks of launch, Notion users created 18,000 Claude agents that completed 140,000 steps, with 90 percent of agent activity running through automated workflows rather than direct chat, while retaining granular permission controls and human oversight. Source

4. Deepgram Reports 4-10x More Durable Code Using Claude

Anthropic published a customer story on Deepgram’s engineering overhaul around Claude. Deepgram found that regular and power Claude users produced 4 to 10 times more durable code than non-users, cut customer incident triage from multi-day investigations to minutes, and reached roughly 95 percent Claude-written code on its most productive team after migrating to Claude Enterprise and building reusable Claude Code skills. Source

5. Claude Code v2.1.239 Adds Fullscreen Renderer for More Platforms and a Python SDK Migration Command

Anthropic tagged Claude Code v2.1.239 on August 21. New additions include a one-time offer to enable the fullscreen renderer on Bedrock, Vertex, Foundry, and other previously excluded setups, with new installs there now starting in fullscreen, plus a /claude-api upgrade command that migrates Python projects from the anthropic SDK 0.x to 1.x. Source

6. Anthropic Hires Former Google TPU Chief as It Builds Custom Chip Effort

Anthropic hired Amir Salek, the founder and former head of Google’s custom silicon program who led development of the first seven generations of Google’s TPUs, to join its compute team. The hire, reported by Bloomberg, comes as Anthropic lays the groundwork for designing its own AI chips and co-designing hardware with its models to improve inference speed and efficiency. Source