CrowdStrike Brings Falcon AIDR Into Claude Cowork
At Fal.Con Europe on Nov. 3, CrowdStrike said it is building Falcon Guardian AI Detection and Response natively into Claude Cowork, developed with Anthropic. It targets malicious agent activity and data exfiltration, with central control through the Falcon platform on Windows and Mac.
This is a paid press release, and it notes the features are still in development. Treat the timing as a roadmap, not a ship date.
Invisible Watermarks Reach Four Older Models
Starting Nov. 3, Claude Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6 carry an invisible text watermark on the Claude Platform, with Amazon Bedrock following a few days later. Newer models such as Opus 5.5 already have it.
The mark is a statistical pattern in word choice. It adds no tokens, costs nothing extra and holds no user data. There is no public checker yet, and a full rewrite removes it. The rollout tracks EU AI Act marking rules.
Message Threads Enter Public Beta
Anthropic's Nov. 2 beta lets agents send the full conversation once, then only new messages plus the ID of Claude's last reply. Anthropic says wait time drops about 35% on Claude Code requests of 500K to 1M tokens.
It saves time, not money. Claude still reads the whole conversation each turn, and prices and limits stay the same. Threads live encrypted in memory, expire five minutes after the last request, and are not eligible for zero data retention or HIPAA setups.
Opus 4.8 Fast Mode Sunset Is Four Weeks Out
Fast mode for Claude Opus 4.8 was deprecated on Oct. 30 and is removed Nov. 30. Move to Opus 5.5 to keep the speedup. Also recent: the beta Usage and Cost Analytics API (Oct. 29) and a Messages API request cap raised from 32 MB to 128 MB (Oct. 28).
Claude Code 2.1.311 made ultrathink run at xhigh effort on several models and fixed freezes from very long CLAUDE.md list items.
Agent Skills Donated to the Agentic AI Foundation
Anthropic's Oct. 22 post says it is donating Agent Skills to the Agentic AI Foundation, moving the format toward neutral, shared governance. It is still the most useful standard to watch if you build reusable skills.
Expect other vendors to adopt it more openly now that no single company owns it.
GSA Partnership Widens Claude Across Government
On Oct. 27 Anthropic announced an expanded partnership with the GSA to bring Claude to all levels of government. Details beyond the headline were not available to us, so check the original post.
Newsroom shows no other post dated Nov. 2 or Nov. 3 at time of writing.
Trust Plumbing Is the Product Now
Watermarks, endpoint security for agents and cheaper long sessions are unglamorous. They are also what lets regulated companies put agents into production. Anthropic filed its S-1 on Oct. 26, and this is the kind of boring reliability public-market buyers reward.
Watch for Bedrock's watermark date and whether other vendors copy the CrowdStrike pattern.