Claude finds CRISPR-like enzyme, Sanders wants to jail AI builders, YouTube goes full AI
Anthropic's wet lab scores a science first, Bernie wants superintelligence developers in prison, and YouTube rewrites the creator playbook with AI.
Subscribe free →Anthropic's Lab Had Claude Go Fishing in a DNA Database — and It Found Something Big
Anthropic's newly-launched wet lab says Claude autonomously discovered a novel enzyme system that the company is comparing to the machinery behind CRISPR gene editing. Claude found it by sifting through a massive DNA sequence database with no human directing each step. This is the lab's first published result, and the timing — right before Anthropic's expected IPO — is not a coincidence.
Why it matters: If the discovery holds up to scrutiny, it's the clearest demonstration yet that frontier AI can do real, novel science rather than just summarize papers.
Bernie Sanders Wants to Make Building Superintelligence a Federal Crime
Sen. Bernie Sanders and Rep. Greg Casar introduced the Ban Artificial Superintelligence Act, which would criminalize the development of ASI — defined as AI capable of "the destruction or disempowerment of humanity" — with penalties up to 20 years in prison. It's the bluntest piece of AI legislation proposed yet by any U.S. lawmaker. The bill has no co-sponsors listed and faces long odds in the current Congress.
Why it matters: Even if it goes nowhere, the bill signals that the political left is done waiting for voluntary industry commitments and is ready to frame AI safety as a criminal justice issue.
YouTube's Made on YouTube Event Was Basically an AI Product Keynote
At its annual Made on YouTube creator event, Google announced a wall of AI features: conversational video editing via chat, AI-generated thumbnails with A/B testing, live dubbing, dynamic thumbnails, Gemini-powered custom feeds, an "Ask YouTube" shopping assistant, and creator idea generation tools. The throughline is that YouTube is increasingly automating the entire optimization loop — from what to make, to how to title it, to who sees it.
Why it matters: YouTube is betting that removing the algorithmic guesswork will keep creators on-platform and attract more of them, but it also raises the question of whether all AI-optimized content eventually looks the same.
ChatGPT Can Now Run Agentic Tasks by Voice on Your Phone
OpenAI rolled out voice-based agentic features to the ChatGPT mobile app for Pro and Plus users. Through a new "Work" tab, users can speak requests and have ChatGPT complete multi-step tasks autonomously — not just answer questions. It's the closest ChatGPT has come to a true voice-controlled AI agent on mobile.
Why it matters: Voice is the interface that makes AI agents accessible to people who won't type long prompts, and this is OpenAI planting a flag before Apple Intelligence and Google Assistant can fully catch up.
Heavy AI Users Are Still Afraid of AI — and They Want Regulation
A new report found that Americans who use AI daily are just as worried about the technology as those who barely touch it. Familiarity isn't breeding comfort: daily users still broadly support government regulation, and their concerns about job displacement, misinformation, and loss of control track closely with the general public's. More usage does not equal more trust.
Why it matters: This demolishes the "people will warm up to AI once they use it more" argument that the industry has been leaning on to resist regulation.
Snorkel AI Triples Its Valuation to $3.5B as Training Data Becomes the Bottleneck
Snorkel AI closed a $350M Series E, tripling its valuation, on the strength of surging enterprise demand for its data-as-a-service platform. The seven-year-old company helps organizations build, curate, and label the training data that AI models actually need to be useful inside a company. With model capabilities racing ahead, the data layer is increasingly where the real constraint lives.
Why it matters: The massive valuation jump is a clear signal that investors now view proprietary training data infrastructure as a foundational layer — not a commodity service.
Quick Hits
- →Spotify launched Taste Profile in the U.S., letting Premium users see and edit their listener profile in natural language — the algorithm, finally made visible. TechCrunch
- →Enveda raised $311M at a $2B valuation to push its AI-discovered nature-derived drugs into clinical trials, including a post-GLP-1 weight-preservation drug. TechCrunch
- →Qualcomm's new flagship smartphone chip can run a 30B mixture-of-experts model locally, making on-device AI significantly more capable. TechCrunch
- →Microsoft disrupted EvilTokens, an AI-assisted cybercrime platform that compromised 12,000 accounts by automating credential attacks end-to-end. Ars Technica
- →California Gov. Newsom signed bills requiring data centers to disclose their electricity and water consumption — a direct response to AI infrastructure's growing environmental footprint. The Verge
AI TLDR