In today’s edition of Data Points, you’ll learn more about:
- Europe’s new media licensing initiative
- ArXiv’s defenses against AI-assisted mistakes
- The Mythos effect on financial institutions
- Ukraine’s evolving drone and data strategy
But first:
Survey shows Anthropic’s enterprise surge
Anthropic passed OpenAI in business adoption for the first time in April, reaching 34.4 percent of businesses while OpenAI fell to 32.3 percent, per the Ramp AI Index. Over the past year, Anthropic quadrupled its adoption rate while OpenAI grew 0.3 percent. But Ramp’s lead economist warns against reading this as a coronation. Three headwinds threaten Anthropic’s position: its business model incentivizes higher token consumption even when cheaper alternatives suffice (Uber already exhausted its 2026 AI budget on Claude); recent service degradation from compute constraints; and a product roadmap that tripled token costs for image-inclusive prompts. Cheaper inference platforms and OpenAI’s Codex are gaining ground among cost-conscious buyers, so the lead may not hold. (Ramp)
Cerebras: the first of many AI IPOs to come
Cerebras, a Silicon Valley AI chipmaker, surged 89 percent on its first trading day, a sign that investor enthusiasm has moved beyond software and into the infrastructure layer of generative AI. The IPO is part of a broader push toward public markets by AI companies including SpaceX, OpenAI, and Anthropic, firms that until recently showed little urgency to list. The timing matters: enterprise AI spending is accelerating, and compute constraints at providers like Anthropic have made raw chip capacity a strategic bottleneck. For investors, Cerebras is a bet not on any single model but on the proposition that demand for inference and training hardware will keep climbing regardless of which lab wins the business market — even as cost-conscious buyers scrutinize what they pay per token. (The New York Times)
European Union pushes for licensing of copyrighted material
The European Commission is preparing legislation to establish clearer licensing processes for creative works, including writing, art, and music, used to train AI models. The move addresses a core tension: models train on vast quantities of copyrighted material, sparking lawsuits and complaints from creators across Europe. The proposed law would complement existing copyright rules by improving transparency around which works are used, establishing mediation and arbitration channels for licensing negotiations, and ensuring creators receive fair compensation. Tech Commissioner Henna Virkkunen framed it as removing friction: many rights holders want to license their work but need assurance they’ll be paid. (Euractiv)
ArXiv cracks down on sloppy AI-assisted papers
ArXiv will ban authors for a year if their submissions contain “incontrovertible evidence” of unvetted AI-generated work. Computer science section chair Thomas Dietterich clarified that authors are responsible for checking AI output for hallucinated references, plagiarism, errors, and misleading content. Incontrovertible evidence means obvious markers of unchecked LLM generation, including meta-comments telling the author to fill in real data, or summary instructions left in the text. After the ban, authors must have a submission accepted at a peer-reviewed venue before ArXiv will host their work again. (404 Media)
Global finance heads to meet with Anthropic to discuss Mythos
Anthropic is meeting with the Financial Stability Board (the global finance watchdog chaired by Bank of England governor Andrew Bailey) to discuss the security implications of its Claude Mythos model, which has shown advanced capability in finding previously unknown software vulnerabilities. Anthropic has withheld Mythos from public release but granted access to selected firms including Apple and JP Morgan to stress-test its findings. The UK’s AI Security Institute found that the latest Mythos iteration was the first AI to solve a previously unsolved cybersecurity challenge called “cooling tower,” succeeding in three of ten attempts, and noted that frontier models’ ability to complete autonomous cyber tasks has roughly doubled in recent months. Some experts argue Mythos represents evolution rather than revolution—most breaches still stem from weak authentication and unpatched vulnerabilities—but the pace of capability gains has prompted regulators and central banks to coordinate oversight. (The Guardian)
Ukraine drones target war infrastructure, yield training data
Ukraine’s 35-year-old defense minister Mykhailo Fedorov is pushing the military to adopt AI-powered autonomous weapons as leverage to force Russia toward settlement. The tech entrepreneur, four months into the role after leading government digitization, believes warfare should be progressively offloaded to machines: “autonomous weapons are the new nuclear weapons.” AI currently assists with drone target recognition; Fedorov envisions unmanned systems fighting entirely without soldiers in the kill zone. His strategy—Air, Land, Economy—aims to intercept 95 percent of incoming Russian strikes, attrit Russian forces faster than they can be replaced, and cripple the Russian economy. Traditional commanders have pushed back, questioning whether the vision disconnects from battlefield reality, though frontline brigades have largely embraced the technology. Fedorov is also monetizing Ukraine’s war data, including over five million annotated drone videos, through Avenger Labs, licensing it to allied companies to train AI models on the condition that Ukraine receives the resulting systems. (The New York Times)
Want to know more about what matters in AI right now?
Read the latest issue of The Batch for in-depth analysis of news and research.
Last week, Andrew talked about the development of AI Andrew, an AI companion shaped by his personality, and his beliefs about respectful, empathetic, and technically precise communication.
“Reflecting on my beliefs about how to communicate has been an interesting exercise. I believe in: Respect for the individual. I hold a lot of respect for pretty much everyone I talk to, at any experience level or stage of life. I hope that comes through whenever I communicate.
Read Andrew’s letter here.
Other top AI news and research covered in depth:
- The U.S. government plans to evaluate upcoming AI models for national security risks and other hazards prior to their release.
- OpenAI is challenging leaders in speech-to-speech technology with RealTime API updates that enhance audio models' reasoning, transcription, and translation capabilities.
- China’s state regulators have blocked the Meta-Manus acquisition, preventing the tech giant from acquiring the agentic startup based in Singapore.
- Google’s breast cancer detection models are being tested under real-world conditions in clinics, as two studies assess their effectiveness in mammogram diagnosis.
A special offer for our community
In case you missed it, DeepLearning.AI launched our first-ever subscription plan for our entire course catalog! As a Pro Member, you’ll immediately enjoy access to:
- Nearly 200 AI short and long courses from Andrew Ng and industry experts
- Labs and quizzes to test your knowledge
- Projects to share with employers
- Certificates to testify to your new skills
- A community to help you advance at the speed of AI
Enroll now to lock in a year of full access for $25 per month paid upfront, or opt for month-to-month payments at just $30 per month. Both payment options begin with a one-week free trial. Explore Pro’s benefits and start building today!
Data Points is produced by human editors with AI assistance.