1. The music didn't leak from Suno — Suno leaked, and its source code showed where the songs came from Suno's training data left the company the way its secrets were always going to: through a breach, not a disclosure.

2. Prompt injection graduated from lab demo to a working data heist, and vendors are answering with AIs that attack themselves A Claude conversation that looks routine can end with a stranger holding your name, your employer, and the answers to your security questions.

3. xAI Walks Into Court as Plaintiff; Meta Walks In as Defendant, Both Because of AI xAI filed suit against a South Carolina man named Terry Wayne Harwood, alleging he used its Grok chatbot to produce child sexual abuse material.

In Brief

• Apple cleared to launch Apple Intelligence in China using Alibaba's Qwen Apple secured Chinese regulatory approval to ship its AI features in the country, powered by Alibaba's Qwen models rather than its own. The deal ends a delay that kept Apple Intelligence off Chinese iPhones.

• Thinking Machines released its first open model, Inkling Thinking Machines published Inkling, its first public release after roughly 18 months building infrastructure privately. The company frames the model around task-specific tuning rather than one general-purpose system.

• Anthropic and Blackstone back Ode to embed engineers inside enterprises Ode launched with Anthropic and Blackstone funding to place forward-deployed engineers inside companies adopting AI. The bet treats implementation and integration, not model access, as the constraint on enterprise deployment.

• OpenAI researcher Miles Wang in talks to raise for a $2B drug discovery startup Miles Wang, an OpenAI researcher, is discussing funding for an AI drug discovery startup at a roughly $2 billion valuation. The talks are early and target applying models to life sciences research.

• Indian coding startup Emergent hit unicorn status with a $130M Series C Emergent raised $130 million just over a year after launch, reaching a $120 million annualized revenue run rate. The AI coding startup reports more than 200,000 paying customers.

• Microsoft patched 570 vulnerabilities in one Patch Tuesday, crediting AI Microsoft fixed 570 security flaws in its monthly release, a company record, attributing many discoveries to AI-assisted analysis. The patches span its product line.

• Vint Cerf is drafting a standard to identify AI agents online Vint Cerf, co-designer of TCP/IP, is developing a protocol to identify autonomous AI agents operating on the public internet. The work aims to let services distinguish agents from human traffic.

• Rime raised $24M Series A for enterprise voice agents Rime closed a $24 million Series A to handle customer phone calls for enterprises. The company says it now processes over 100 million calls per month across its clients.

• Whatnot acquired Shaped for real-time shopping recommendations Livestream shopping platform Whatnot bought Shaped, a machine learning startup focused on real-time recommendation and search. Whatnot plans to use it for personalization as it adds product categories.

• OpenAI shipped a $230 light-up keyboard for its Codex coding app OpenAI released a $230 keyboard designed to pair with Codex, its agentic coding tool. The launch comes while OpenAI defends against Apple's trade secret theft lawsuit over hardware.

• Anthropic committed $10M to Canadian AI research Anthropic pledged $10 million to fund AI research in Canada. The commitment supports academic and research institutions in the country.

Read the full edition →