Prompt Notes #4 - weekly AI digest

AI goes fast, helping you (and myself) catching up.

Welcome to the 4th edition of Prompt Notes, your new favourite weekly AI digest.

This edition will cover the week from 23 February to 1 March.

What caught my attention

Anthropic holds its red lines against the Pentagon. Trump bans all federal use of their tech. Hegseth designates them as a supply chain risk. Sam Altman publicly praises Anthropic’s stance on autonomous weapons and mass surveillance.

Then signs the Pentagon contract himself hours later.

Incredible scenes.

(Yes, I know, 2 weeks in a row where Sam Altman catches my attention, but damn…)

---

TLDR

• Trump ordered all federal agencies to immediately cease using Anthropic’s technology. Hegseth designated the company a “supply chain risk,” directing all military contractors to cut commercial ties with Anthropic. (CBS News)

• Anthropic refused to allow “all lawful” military use of Claude despite losing its $200 million Pentagon contract, with CEO Dario Amodei stating the company “cannot in good conscience accede.” Autonomous weapons and mass surveillance remain non-negotiables. (The Hill)

• OpenAI’s Sam Altman publicly praised Anthropic’s red lines on autonomous weapons and mass surveillance - then signed the Pentagon contract hours after Anthropic was blacklisted. (OPB)

• Pentagon CTO Emil Michael called Amodei a “liar” with a “God-complex” on X. Amodei called the designation “unprecedented” and “retaliatory” and vowed to challenge it in court. (CBS News)

• Legal experts argue Hegseth’s supply chain risk designation exceeds statutory authority and will not survive judicial review. (Lawfare)

---

Rest of the news

• Researchers from ETH Zurich and Anthropic demonstrate that commercially available AI models can identify pseudonymous internet users for a few dollars per person. (The Decoder)

• OpenAI published details of its contract with the Department of War, including safety red lines, legal protections, and deployment in classified environments. (OpenAI Blog)

• ElevenLabs and Google lead the updated Artificial Analysis speech-to-text benchmark. (The Decoder)

• Anthropic’s Claude rose to No. 1 in the App Store amid its Pentagon negotiations. (TechCrunch AI)

• Block Inc. laid off nearly half its staff, with co-founder Jack Dorsey attributing it to AI enabling more work with fewer employees. (Bloomberg Technology)

• China’s Honor debuted a smartphone with a robotic camera arm at Mobile World Congress and teased a humanoid robot. (CNBC Tech)

• ZTE showcased full-stack AI innovations under its “Connectivity + Computing” strategy at MWC Barcelona 2026. (Capacity Global)

• AI experts warn that complex AI systems risk surpassing human comprehension, potentially leading to business chaos. (CNBC Tech)

• AI servers now dominate Dell’s systems business. (Next Platform)

• Frontier LLMs like GPT-5.2 and Claude 4.6 lose up to 33% accuracy in longer conversations. (The Decoder)

• OpenAI labeled AI expert Stuart Russell a “doomer” in court despite CEO Sam Altman previously co-signing his AI extinction warnings. (The Decoder)

• OpenAI promised Canada tighter safety protocols after ChatGPT flagged a shooter’s violent chats but did not alert police. (The Decoder)

• AI advancements lack sufficient guardrails amid market meltdowns and political fights. (CNBC Tech)

• Health officials in Illinois used an AI chatbot to investigate a puzzling outbreak linked to a county fair. (Ars Technica AI)

• Google integrated its Intrinsic robotics project into the main company to develop an “Android for robotics” platform. (CNBC Tech)

• CoreWeave CEO Mike Intrator defended accelerated infrastructure spending amid an 18% stock drop. (CNBC Tech)

• AI music generator Suno reached 2 million paid subscribers and $300 million in annual recurring revenue. (TechCrunch AI)

• Accenture partnered with Mistral AI to scale secure, sovereign AI deployments across Europe. (AI Business)

• Anthropic’s Claude Code now automatically remembers debugging patterns, project context, and user preferences across sessions. (The Decoder)

• Bing updated its webmaster guidelines to include GEO for Copilot grounding, meta directives for AI answers, and expanded AI abuse definitions. (SEJ Generative AI)

• Elon Musk criticized OpenAI in a deposition, claiming xAI’s Grok caused no suicides unlike ChatGPT. (TechCrunch AI)

• OpenAI updated its mental health safety measures, including parental controls, trusted contacts, and distress detection. (OpenAI Blog)

• eBay laid off about 800 roles, or 6% of its workforce, to focus on strategic priorities. (CNBC Tech)

• Alibaba released the open Qwen 3.5 model series, including Qwen3.5-Flash, Qwen3.5-35B-A3B, Qwen3.5-122B-A10B, and Qwen3.5-27B, targeting GPT-5 Mini and Claude Sonnet 4.5 at a fraction of the cost. (The Decoder)

• Bumble added AI-powered photo feedback and profile guidance tools. (TechCrunch AI)

• Burger King launched AI chatbot “Patty” in employee headsets to assist with meal prep and evaluate customer interaction friendliness. (The Verge AI)

• Mark Zuckerberg’s appearance at Prada’s fashion week sparked speculation about Prada-branded Meta AI glasses. (TechCrunch AI)

• C3 AI cut 26% of its workforce and reported wider-than-expected losses under new CEO Ehikian’s restructuring. (CNBC Tech)

• Adobe’s Firefly introduced a “Quick Cut” tool to generate rough edits from raw footage via text prompts. (The Decoder)

• The Trump administration directed U.S. diplomats to lobby against foreign data sovereignty laws affecting American tech firms. (TechCrunch AI)

• David Luan, head of Amazon’s San Francisco AGI lab, announced his departure after less than two years. (The Verge AI)

• The White House urged AI companies to cover electricity rate hikes, with many hyperscalers already committed. (TechCrunch AI)

• The FBI is investigating X over nonconsensual nude images generated by Grok AI. (Social Media Today)

• Google’s Gemini on Android now automates multi-step tasks like rideshare requests and grocery deliveries. (TechCrunch AI)

• Anthropic highlighted risks of large-scale model distillation by Chinese vendors for enterprises. (AI Business)

• The $500 billion Stargate AI data center project stalled due to disputes among OpenAI, Oracle, and SoftBank. (The Decoder)

• Multiple AIs were tested for parsing 20,000 pages of Jeffrey Epstein documents released by the House Oversight Committee. (The Verge AI)

• Spotify rolled out AI-powered Prompted Playlists to Premium subscribers in the U.K., Ireland, Australia, and Sweden. (TechCrunch AI)

• Google released Nano Banana 2, leveraging Gemini’s knowledge base for real-time web search information and images. (AI Business)

• Hugging Face published a blog on Mixture of Experts (MoEs) in Transformers. (Hugging Face Blog)

• Stratechery interviewed VC Bill Gurley on his book, Uber, and the VC industry. (Stratechery)

• VectifyAI launched Mafin 2.5 and open-source PageIndex, achieving 98.7% accuracy in financial RAG via vectorless tree indexing. (MarkTechPost)

Thanks for reading! Subscribe for free to receive new posts.