Anthropic released Claude Code v2.1.148, a patch fixing a critical Bash tool regression introduced in v2.1.147 that caused every command to return exit code 127 for some users. The hotfix restores normal terminal execution inside Claude Code sessions.
Update via the built-in auto-update or by reinstalling.
OpenAI launched 'OpenAI for Singapore' with MDDI at ATxSummit, committing over S$300M (US$234M) to open its first applied AI lab outside the US. The lab scales Singapore-based technical staff to 200+ roles and serves as a global hub for Forward-Deployed Engineers, focused on public service, finance, healthcare and digital infrastructure.
Includes an OpenAI Academy chapter and AI learning programs with MOE and GovTech.
Google DeepMind announced an expanded National AI Partnership with Singapore at ATxSummit 2026. The MOU with MDDI scales frontier AI across public services, health, and education — including AI co-clinician pilots, AlphaFold-powered pandemic preparedness, a Gemma-based running assistant for low-vision athletes, Gemini for Education, an 'AI for the Planet' accelerator, and multilingual safety benchmarks with IMDA and MLCommons.
Google launches 'A National Portrait' with artist Es Devlin at the UK's National Portrait Gallery. UK residents 18+ can upload photos via Google Arts & Culture, where the Gemini Image model transforms them into animated charcoal-style portraits styled after Devlin's drawings.
Participants opt in to join an evolving collective artwork on display through October 27, 2026.
Microsoft rolls out the Windows 11 May 2026 update with Xbox mode for a console-like PC experience, expanded Taskbar AI agent monitoring, faster File Explorer with broader archive support, stricter driver protocols, and tighter security defaults. The update lands today across Windows 11 PCs, laptops, desktops, and tablets.
Google Cloud outlines its agentic-era public sector blueprint: eighth-generation TPUs in the AI Hypercomputer, Gemini and the full Google AI stack via Google Distributed Cloud, an agentic data cloud with Knowledge Catalog and a cross-cloud Lakehouse, plus expanded threat detection with Cloud Armor and Model Armor for governments scaling AI from pilots to production.
OpenAI launches Campus Network, a global initiative connecting student clubs at universities with AI tools, event support, and frontier-model access. Clubs can apply via an interest form to host workshops, hackathons, and community events on their campuses.
The program is OpenAI's first structured push to build student-led AI communities and grow its presence in higher education.
Claude Code v2.1.137 and v2.1.138 released on May 9. v2.1.137 fixes VS Code extension failing to activate on Windows.
v2.1.138 ships internal stability fixes. Both are maintenance releases following the major v2.1.136 drop yesterday (enterprise feedback survey support, stricter auto mode rules, MCP/OAuth improvements, extended thinking fixes, terminal rendering).
Upstage AI publishes a guide on building modular AI tech stacks for insurance and document-heavy industries. The article argues that monolithic end-to-end platforms fail under real-world variability and advocates for decoupled architectures: document parsing → extraction → model → output → downstream systems.
Key insight: swappable model layers prevent costly overhauls as AI evolves rapidly.
Anthropic hosts a free live webinar today: "Claude for Nonprofits: Moving Your Workflow Beyond Chat" at 10 AM PT. The session shows the same nonprofit workflow (grant writing) run through Chat, a Project, a Skill, and a Connector — live side-by-side — so organizations can see exactly how to move beyond basic chat for higher-impact work.
https://www.anthropic.com/webinars/claude-for-nonprofits-moving-your-workflow-beyond-chat
Claude Code v2.1.129 and v2.1.131 released today. v2.1.129 adds --plugin-url flag for fetching plugin .zip archives from URLs, CLAUDE_CODE_PACKAGE_MANAGER_AUTO_UPDATE for background upgrades, gateway model discovery opt-in, plus multiple bug fixes for /clear, /rename, MCP, and emoji rendering. v2.1.131 fixes VS Code extension activation on Windows and Mantle endpoint auth.
https://code.claude.com/docs/en/changelog
Anthropic's Code with Claude developer conference takes place today in San Francisco. A full day of hands-on workshops, live demos of new Claude capabilities, and office hours with Anthropic engineers — covering agentic coding, MCP ecosystem integrations, and production reliability. Keynote speakers include Ami Vora, Boris Cherny, and Angela Jiang. Virtual livestream available globally.
https://evermx.com
OpenAI enabled marketing cookies by default for free ChatGPT users, sharing limited identifiers (cookie IDs, email addresses) with ad partners to promote OpenAI products across third-party sites like Instagram. Paid subscribers (Plus, Enterprise) are exempt. Users can opt out via Settings > Data Controls > Marketing Privacy.
https://openai.com/policies/cookie-policy/
xAI drops Grok 4.3 with ~40% input price cut ($1.25/M tokens) and new Imagine Agent mode for creative projects. The model scores 53 on the AI Intelligence Index with a 1M-token context window. Imagine Agent enables autonomous multi-step creative workflows — generate a full film, manga set, or product photoshoot from a single prompt.
https://x.ai/news
Alibaba Qwen releases Qwen-Scope, an open-source suite of Sparse Autoencoders (SAEs) for the Qwen3 and Qwen3.5 model family. 14 SAE groups enable feature steering (no prompt engineering), dataset classification, and interpretability analysis.
Models range from 2B to 35B parameters. Try it via QwenScope Space on HuggingFace.
Microsoft Agent 365 and Microsoft 365 E7 are now generally available. Agent 365 ($15/user/mo) is the enterprise AI control plane for governing and securing agents across the org.
M365 E7 ($99/user/mo) bundles E5 + Copilot + Entra Suite + Agent 365 — Microsoft's first new enterprise tier since E5 in 2015.
Anthropic launches Claude Security in public beta for Enterprise customers. Powered by Claude Opus 4.7, it scans codebases for vulnerabilities and generates targeted patches — no API integration required.
Partners include CrowdStrike, Palo Alto Networks, SentinelOne, Wiz, Accenture, and Deloitte. Team and Max plans coming soon.
Anthropic retires 1M token context window beta for Claude Sonnet 4.5 and Claude Sonnet 4 today (Apr 30). The `context-1m-2025-08-07` beta header no longer works on these models; requests exceeding 200k tokens will return an error.
Migrate to Sonnet 4.6 or Opus 4.6 for full 1M context at standard pricing.
Google DeepMind announced a strategic partnership with South Korea's Ministry of Science and ICT to advance AI-driven scientific discovery. An AI Campus will be set up in Seoul, linking top institutions like Seoul National University and KAIST with Google's frontier models — AlphaFold, AlphaGenome, and WeatherNext.
Korean students gain internship access, and Korea's AI Safety Institute joins on research best practices. This marks 10 years since the historic AlphaGo match in Seoul.
https://deepmind.google/blog/announcing-our-partnership-with-the-republic-of-korea/
Google Cloud Next '26 Day 2 developer keynote unveiled the full Gemini Enterprise Agent Platform — the evolution of Vertex AI. Key launches: Agent Development Kit (ADK), Agent-to-Agent Protocol (A2A), Agent-to-User Interface (A2UI), Agent Registry ("DNS of your internet of agents"), and Agent Runtime with memory. MCP server support and Wiz Green Agent security scanning also announced.
https://cloud.google.com/blog/topics/google-cloud-next/next26-day-2-recap
