METR Finds Anthropic AI R&D Pace Has Nearly Doubled
METR’s 5 November 2026 assessment finds Anthropic likely improving AI capabilities ~2x faster since late 2025, mostly via AI used for R&D.
LTM, ServiceNow Expand Agentic AI Partnership for 700 Clients
LTM and ServiceNow expand their agentic AI partnership for 700-plus joint customers, using LTM’s own 1,500 ServiceNow AI agents as the blueprint.
TCS Enterprise AI Study: Only 10% Let AI Agents Act Alone
TCS Enterprise AI Study finds 84% of leaders see productivity as AI’s top gain, but only 10% let AI agents act without human approval across functions.
CERT-In Issues AI Security Audit Rules With 6-Hour Reporting
CERT-In has issued AI security audit guidelines that sort AI systems into four risk tiers and require AI incidents to be reported within six hours.
CERT-In Gets Access to Anthropic’s Mythos AI for Cyber Defence
CERT-In has gained access to Anthropic’s unreleased Mythos AI model to find and fix system vulnerabilities, starting slowly over sensitive-data concerns.
Straive Partners Anthropic to Run Responsible Enterprise AI
Straive and Anthropic partnered on 3 November 2026 to help enterprises build, deploy and run AI responsibly, with governance and human oversight built in.
Salesforce Gives CIOs a Kill Switch for AI Agents, Adds Live Evals and Hard Spending Caps to Agent Fabric
Salesforce has added Agent Evals, API Evals, an Agent Kill Switch and hard budget limits to Agent Fabric, its cross-platform control plane for AI agents, so enterprises can catch failing or costly agents and take them offline before customers notice.
Trojanised Claude Code Update on Cursor’s Marketplace Tried to Plant SSH Backdoors on Up to 5,000 Developer Machines
A tampered Claude Code extension was served through Cursor’s marketplace for nearly 11 hours on 29 October, attempting to install a persistent SSH backdoor on up to 5,000 developer machines, a warning for enterprises scaling AI coding tools.
Quad Pledges Over $85 Million to Widen Access to Trusted AI Across the Indo-Pacific, Will Map Chip and AI Hardware Chokepoints
Meeting in Atlanta on 31 October, the foreign ministers of India, the US, Japan and Australia said they intend to provide more than $85 million in technical and financial assistance to support access to trusted AI across the Indo-Pacific, and will launch an effort to spot semiconductor and AI hardware supply-chain chokepoints before they cause disruption.
OpenAI, Google, Microsoft and Anthropic Back One Opt-Out Signal for AI Training, but Indian Publishers Want More
OpenAI, Google, Microsoft and Anthropic will honour a single robots.txt signal that lets websites opt out of generative AI training, while Microsoft adds a separate grounding control. Indian publishers say it does not address content already used.
