📖 2 min read
Your daily briefing on the biggest moves in artificial intelligence. Here’s what happened in the last 24 hours.
Anthropic Unveils Project Glasswing for AI-Powered Cybersecurity
Anthropic announced Project Glasswing, a defensive cybersecurity initiative built on its unreleased Claude Mythos Preview model. The project gives 11 select organizations – including Google, Microsoft, and JPMorgan Chase – limited access to what Anthropic calls the most capable AI system ever built for finding and fixing software vulnerabilities.
📧 Want more like this? Get our free The Ultimate AI Tool Database: 200+ Tools Rated & Ranked — Downloaded 5,000+ times
The model wasn’t trained specifically for cybersecurity. Anthropic says the capabilities emerged naturally from its strong agentic coding and reasoning skills. The company is keeping the release tightly controlled while testing new cyber safeguards on less capable models first.
Mozilla Launches Thunderbolt, an Open-Source AI Client
Mozilla’s for-profit arm MZLA Technologies just open-sourced Thunderbolt – a full AI workspace that runs on your own hardware, supports any model (local or cloud), and works across platforms. It’s available now on GitHub.
The pitch is simple: a front-end client for users and businesses who want to run their own self-hosted AI infrastructure without relying on cloud-based third-party services. If you’ve been looking for an open alternative to locked-in AI platforms, this one’s worth checking out.
📧 Want more like this? Get our free The Ultimate AI Tool Database: 200+ Tools Rated & Ranked — Downloaded 5,000+ times
Claude Design Rolls Out – AI Prototyping for Non-Designers
Anthropic launched Claude Design, an experimental AI tool that creates prototypes, pitch decks, slides, and one-pagers through conversational prompts. Powered by the new Claude Opus 4.7, it’s rolling out as a research preview under Anthropic Labs.
The tool targets non-designers who need professional-looking visual output fast. Think startup founders building investor decks or product managers mocking up UI flows – no Figma skills required.
EU Locks In AI Hiring Bias Rules – 105 Days to Comply
The EU AI Act’s employment provisions are now 105 days from enforcement. Any AI tool used to evaluate EU-based candidates – resume-ranking algorithms, AI interview scoring, job ad targeting – qualifies as high-risk and must comply.
📧 Want more like this? Get our free The Ultimate AI Tool Database: 200+ Tools Rated & Ranked — Downloaded 5,000+ times
What that means in practice: annual third-party bias audits, full technical documentation, human oversight mechanisms, and transparency disclosures to candidates. Non-compliance penalties hit hard – up to 15 million euros or 3% of global annual turnover.
Sam Altman’s Outside Ventures Raise Conflict-of-Interest Questions
Investigative reports published today detail how OpenAI CEO Sam Altman’s personal investment portfolio – spanning nuclear energy startup Oklo, chip ventures, and other AI-adjacent companies – is raising questions about who OpenAI is really building for.
The scrutiny comes as OpenAI continues its transition from nonprofit to for-profit, and as GPT-6 remains conspicuously absent past its rumored April 14 launch window. Meanwhile, the competition isn’t waiting around.