🤖 White House rolls out voluntary AI safety framework 🤖
A new regulation framework could give the U.S. government early access to frontier AI models.

The White House has invited OpenAI, Anthropic, Meta, and Google to review a newly completed voluntary cybersecurity testing framework, developed under President Trump's June 2 executive order. Under the plan, companies could voluntarily grant the U.S. government access to frontier models up to 30 days before public release. The upcoming meeting will cover a classified benchmark, implementation steps, and open questions — including what qualifies as "frontier AI," whether open-weight models are included, and which entity will lead testing. This is one of the most concrete U.S. moves toward structured pre-release evaluation, shaping how major labs handle safety and compliance. The initiative arrives as the EU AI Act takes effect and over 1,200 AI staffers have publicly called for more measured pacing of frontier development.

🧬 Ex-OpenAI exec launches ChronicleBio for AI drug discovery 🧬
ChronicleBio rolls out an AI-first approach to chronic disease treatment and drug trials.

Fidji Simo, who left OpenAI's leadership team after a seven-year personal battle with Postural Orthostatic Tachycardia Syndrome (POTS), has launched ChronicleBio — a startup dedicated to using AI to cure POTS and other chronic diseases. The company has already amassed 153 terabytes of biological data from blood draws and plans to expand its dataset through home collection services. ChronicleBio's strategy centers on leveraging this data to improve clinical drug trial success rates, signaling growing convergence of top AI expertise and precision medicine. The venture combines deep technical know-how from one of the industry's most prominent labs with a deeply personal health mission, making it one of the most closely watched new entrants in AI-driven drug discovery.

💼 Anthropic CEO: top talent joins for pay, not mission 💼
Anthropic's CEO flags a growing tension between competitive pricing and mission alignment.

Anthropic CEO Dario Amodei publicly voiced concern that new hires are increasingly joining the company for its industry-leading compensation rather than its safety-focused mission. Anthropic reportedly pays more than any other AI lab, yet retaining top researchers remains a persistent challenge as OpenAI, Google, and Meta aggressively recruit and counter-offer. Amodei's remarks underscore a broader industry tension: when pay alone can't differentiate employers, mission alignment, compute access, and research freedom become the real battlegrounds. This raises questions about whether the current AI compensation arms race is sustainable, and how mission-driven organizations maintain their culture at scale.

INTERESTING TO KNOW

📱 Microsoft open-sources Orchard for agentic AI development 📱

Microsoft has released Orchard, an open-source agentic modeling framework built on a thin, Kubernetes-native environment that enables reusable recipes for trajectory distillation, on-policy reinforcement learning rollouts, and evaluations — all on one substrate. Its portability means datasets, training recipes, and evaluation protocols can be shared across projects rather than rebuilt each time, lowering the barrier for teams building agentic AI systems. Available now on GitHub, Orchard positions Microsoft as a provider of foundational open-source tooling alongside its commercial Copilot offerings.

🔌 Google builds plugin system for Gemini Enterprise 🔌

Google is developing a plugin system for Gemini Enterprise that would enable mini-apps and packaged workflows, expanding its integration capabilities for business users. The rollout mirrors how ChatGPT introduced plugins and how Microsoft connects Copilot to its productivity suite. For enterprise customers, plugins could extend Gemini beyond chat into structured business processes and domain-specific applications — intensifying the enterprise AI platform war with Microsoft, where ecosystem depth is becoming the key differentiator.