This week OpenAI announced developments that illustrate a shift from models that answer questions to models expected to perform substantive work. GPT‑6 Astra was described as capable of navigating software interfaces, completing multistep workflows, and applying advanced mathematical and scientific reasoning. OpenAI also introduced ChatGPT for financial services, developed with input from Morgan Stanley and Evercore to support research, financial modeling and the production of client materials.
OpenAI said it had produced a solution to the Navier–Stokes Millennium Prize Problem. According to the company, a coordinated system of 10,000 AI agents worked for 88 hours on the proof, followed by 17 hours of model‑based verification by Astra. External researchers questioned whether OpenAI may have had access to related prior work; OpenAI denies that allegation. While such scientific claims are notable, they prompt questions about how well these systems are understood and what human role should remain in scientific discovery.
Investors betting across the AI stack
Capital continues to flow to AI companies, but investors are spreading bets across infrastructure, platforms and specialized applications rather than concentrating on a single layer. French company Mistral raised €3 billion with plans to invest in compute infrastructure and open‑weight models. Other large financings were reported for the legal AI company Harvey, the inference‑chip startup Positron, and the enterprise AI firm Wonderful. NVIDIA announced the acquisition of Hugging Face for nearly $13 billion.
Data cited by host Christina Stathopoulos showed global AI funding rising from $56 billion in the fourth quarter of 2025 to $242 billion in the first quarter of 2026. She questioned whether generative AI will deliver returns that justify that scale of investment; some companies may build durable businesses and others may not, even if AI continues to produce useful products and services.
Safety concerns collide with high‑value applications
AI safety returned to the spotlight after the high‑profile resignation of Anthropic researcher Jacob Coxon, who warned that labs were moving too quickly toward poorly understood systems capable of recursive self‑improvement. Similar concerns were raised by other researchers affiliated with Anthropic and Google DeepMind. Anthropic CEO Dario Amodei called for stronger evaluation, shared safety standards and international coordination; Sam Altman and Elon Musk seconded that call.
While the industry remains divided over catastrophic‑risk scenarios, many nearer‑term problems are already tangible. Host Christina Stathopoulos expressed greater concern about malicious human use of powerful AI systems than about autonomous systems becoming dangerous on their own. Organizations deploying more autonomous systems should proceed cautiously, implementing robust security, access controls, testing and human oversight.
"AI for good" shows promise in genomics
Balancing the safety debate, the episode highlighted practical AI benefits in genomics. DeepMind, UC Berkeley and Tempus are using AI to predict how genetic changes affect gene function, identify mutations associated with disease, and link genomic data with patients’ medical histories.
For researchers, AI can make it feasible to study genetic possibilities that would be difficult to test individually in the lab. That capability may narrow the search for disease‑related variants and support earlier diagnosis and more personalized treatment.
What’s next
This Week in AI will continue to cover the news, issues and developments shaping the AI era in its next episode; new installments appear each Friday and are available on YouTube, Spotify, Apple and other platforms.
The newsletter also invited readers who work in cybersecurity to answer an 11‑question survey to inform a forthcoming report.
