Anthropic Launches Its Cyber Mission
Anthropic announced a long-term cyber push with two parts. The Critical Infrastructure Defense Program gives partners frontier Claude models, on-site engineers and threat research to protect power grids, water utilities, factories and transport. The OSS Scanner is a free, opt-in service that scans enrolled open-source projects and returns bug reports with a proof-of-concept exploit and a suggested fix.
Anthropic expects a true-positive rate above 90%, but reports go out without human review, so expect some errors. Eleven founding partners: Accenture, Booz Allen, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC and Rockwell Automation.
Usage Policy Bans Sustained Cruelty Toward Claude
The updated policy prohibits sustained, needless abusive behavior toward Claude and election campaigns built on deception, effective Nov 12. Examples include flooding the system with abuse or provoking it after a task ends.
There is no automatic suspension. Claude can end a persistently hostile chat on its own. Constructive criticism, frustration over errors, dark fiction and security testing remain allowed.
Claude for Startups Gets a Bigger Offer
Announced at SF Tech Week: a free year of Claude Team with up to five premium seats, $1,000 in API credits, Claude Marketplace access and virtual office hours with the Applied AI team. Companies founded in the last five years or funded in the last two can apply.
Claude Code 2.1.293 to 2.1.295
Version 2.1.295 adds onFailure: "block" for command and HTTP hooks, so a hook that fails or times out blocks the action. MCP tool descriptions loaded via tool search now cut off at 16,384 characters, up from 2,048. About 97 bugs fixed.
Version 2.1.294 fixes prompt and agent hooks written as instructions letting through actions they should block. Version 2.1.293 made Haiku 5.5 the default Haiku model.
Compliance API Now Covers Unified Claude Chats
The Compliance API chat endpoints now also return chats from the unified Claude experience. It is in beta for Enterprise orgs and uses your existing Compliance Access Key.
Ping Labs Ships a Claude-Built Rust TypeScript Compiler
Theo Browne's ts-rust ports the TypeScript 7 compiler to Rust and passes all 181,711 ported Go compiler tests. Claude Opus wrote a v0 in 10 hours and the final run cost about $24K over two weeks. An earlier Codex attempt reportedly cost over $420K and reached about 84% compatibility.
The README claims 1.61x faster than tsc 7 on average. Known issues include memory growth in long editor sessions and possible stale output in multi-project builds.
Anthropic Reportedly Builds a 2028 Election Program
Fortune reports a presidential engagement program meant to engage candidates from both parties on AI policy. Anthropic is hiring a Political Programs lead at $295K to $345K who needs 12+ years of campaign experience.
SWE-Game: Opus 5 Leads but Stays Under 60/100
The new SWE-Game benchmark covers 247 Godot build tasks. Claude Opus 5 tops the scores, yet still lands below 60 out of 100.
A $99 Smart Ring Routes Voice to Claude
Natura's Interface ring sends voice input to several assistants, Claude among them, alongside ChatGPT and others.
Cyber Is Now a Distribution Problem
Free scanning for open source plus OT defense through big vendors shows Anthropic wants its models in defenders' hands at scale. The catch: routing through incumbents may leave smaller operators with less access.
ts-rust makes the other point. A $24K Opus run beat a $400K+ Codex attempt, so agent work is becoming a budget line item you can plan around.