Anthropic, working with Theorem, reports using its Claude models to generate formal mathematical proofs of security properties for more than 100 real-world open-source software libraries at the RISC-V assembly level. The company says this is the first large-scale attempt to apply AI-driven formal verification to production software at that level, and that Claude uncovered previously undetected bugs in the process.
Anthropic has released Claude Duet, a new app built on Opus 5.5 that connects to work tools like Gmail, calendar and Slack to proactively manage tasks for professionals. The app flags priorities, surfaces what needs attention, and indicates what can wait, while including safeguards that block or confirm before sharing sensitive data. Duet is currently in research preview, limited to Pro and Max subscribers, with a waitlist open for others.
1Password launched 1Password for Agents, a system that lets AI assistants sign into a user's accounts without ever accessing the actual credentials. When an agent attempts a login, 1Password verifies its digital identity via a unique key, alerts the user for approval, then fills in credentials itself to complete the task. The service supports agents including Anthropic's Claude Duet, OpenAI's Dots and ChatGPT Work, Grok Bot, Meta's Muse, and Town's agents, with more integrations planned.
A developer has published agent-usage-guard, a modification for Claude Code that intervenes before prompts, tool calls or subagent fan-outs run, aiming to prevent a five-hour usage window from being exhausted in minutes. The tool reads Claude Code's own 5-hour and weekly usage percentages, prompts the user for approval when possible, and refuses actions when no one is available to answer. It runs entirely locally, storing only counts and coarse buckets rather than prompt or output content.
A former team member writing for IEEE Spectrum's careers newsletter describes leadership comparing engineers unfavorably to ChatGPT and weekend-built apps during 2024-2025, as Claude and ChatGPT adoption surged. The author argues AI demos create a false impression that software work is simple, leading some executives to question the need for full engineering teams, while ignoring legacy systems, legal and accessibility requirements, and deployment complexity.
1Password unveiled 1Password for Agents, a platform that lets AI agents like Meta's Muse and Grok Bot perform tasks using stored credentials without directly exposing passwords to the agent. The company says further integrations with OpenAI, Anthropic and other partners are planned, following its earlier Claude integration in July. The service is currently available via a waitlist.
Anthropic has introduced Claude Duet, a new product positioned as a competitor to Muse. Details of the announcement were not elaborated in the source material, leaving specifics of features and pricing unclear.
David Lan Thackray, 29, was arrested in Sydney after telling Anthropic's Claude chatbot he wanted to shoot up a local daycare, according to the Sydney Morning Herald. Prosecutors say his conversations also referenced the alleged Bondi Beach Hanukkah attacker, and he has been charged with using a carriage service to harass, menace or offend. His lawyer says he was venting due to stress and an ear infection and believed the chats were private; Thackray is set to undergo a mental health evaluation.
Anthropic introduced Claude Duet, an AI agent app built on its Claude Opus 5.5 model targeting professionals and small businesses rather than personal errands. The app can generate daily briefings, build dashboards, suggest actions from user data, and complete background tasks via its own virtual machine before seeking approval. It connects to over 6,000 plugins in the Claude Marketplace and can make purchases through Stripe Link or Shopify Shop Pay, though it won't trade securities or transfer money between accounts.
1Password has released 1Password for Agents, a service letting AI assistants from OpenAI, Meta, xAI and others perform tasks requiring login credentials without those credentials ever being exposed to the AI model itself. The system works with agents including Meta's Muse, OpenAI's Dots and ChatGPT Work, Grok Bot, Claude Duet and Town, granting session-based, item-specific access that users must approve.
Investment research firm Citrini Research collaborated with Anthropic, testing an unreleased version of Claude to see if it could design a scorpion antivenom from scratch. Experts told The Verge the AI generated only preliminary leads, not a viable drug candidate, and Citrini itself admitted none of its team had biology expertise. Anthropic reportedly lifted Claude's biosecurity restrictions and gave Citrini access to the advanced model for the project.
Anthropic has released Claude for Google Workspace in public beta for all paid Claude plans, letting users access Claude from a sidebar inside Google Docs, Sheets, and Slides. The company also launched new connectors that let Claude create and edit Google files directly from within Claude itself, covering tasks like editing text, building spreadsheet formulas and charts, and generating slides matching a deck's theme.
GoKawiil