Musk concludes contentious testimony - The New York Times

Elon Musk’s 7 biggest stumbles on the stand - Ars Technica

Elon Musk sought majority control of OpenAI in 2017, withheld funding, and tried to poach researchers, according to evidence in his trial against Sam Altman. - WIRED

OpenAI lawyers claim Shivon Zilis, mother of four of Elon Musk's children, acted as a confidential liaison between Musk and company leadership. - WIRED

Elon Musk testified that xAI used OpenAI models through distillation to train Grok, defending the practice as a standard industry method. - The Verge

OpenAI President Greg Brockman reports AI tools now write 80% of code, up from 20% in December, while Google reports 75% AI-generated code. - Business Insider

OpenAI CFO reports a ‘vertical wall’ of demand for the company's products. - Bloomberg

How the Microsoft-OpenAI conscious uncoupling will work - The Verge

OpenAI's GPT-5.5 achieved a 71.4% pass rate on cyber tasks and completed a 32-step autonomous attack simulation in UK AISI evaluations. - AI Security Institute

GPT-5.5 completed a 12-hour human cyber-attack task in 11 minutes for $1.73, outperforming Anthropic's Mythos in UK AI Safety Institute evaluations. - Reddit

David Sacks downplays Mythos, notes GPT-5.5 not as compute-constrained. - X (formerly Twitter)

An OpenAI model achieved 67% diagnostic accuracy in emergency rooms, outperforming physicians' 55% rate according to a recent study. - Semafor

Anthropic overtakes OpenAI in Q1 2026 LLM revenue with 31.4% market share and $16.20 per user vs. $2.20 - The Register

Anthropic launched Claude Security in public beta for Enterprise customers, utilizing Claude Opus 4.7 to scan code for vulnerabilities and automate patch workflows. - Implicator.ai

Anthropic's Claude Code tool refuses requests or applies extra charges if "OpenClaw" is mentioned in commit history, according to developer reports. - X (formerly Twitter)

After ghosting Anthropic, the Pentagon signed agreements with six technology companies to expand classified artificial intelligence work. - The New York Times

Pentagon CTO Emil Michael says Anthropic remains blacklisted for supply chain risks as the DOD signs AI agreements with seven companies for classified networks. - CNBC

No single AI source of truth: US officials draft an AI policy memo requiring national security agencies to avoid reliance on single vendors. - Bloomberg

The NSA tests Anthropic's Mythos model to identify vulnerabilities in Microsoft products and other widely used software. - Bloomberg

Defense Secretary Pete Hegseth announced a new autonomous warfare command in the $1.5 trillion 2027 budget, as Human Rights Watch warned of civilian risks. - Antiwar.com

Hegseth defends the Pentagon's use of artificial intelligence and calls the Anthropic CEO a "lunatic." - Bloomberg

gee I wonder why Anthropic wouldn’t want the reputational risk of associating with it.

AI crime wave unlocked: Automated threats are projected to skyrocket global cybercrime costs, following a 368% increase in median ransomware payments. - Financial Times

Alphabet soup of capex: Cloud growth outpaces Amazon and Microsoft as Big Tech companies project $725 billion in AI infrastructure spending. - Financial Times

Nvidia DGX B300 servers available in China at double US price despite being officially unavailable under US export controls. - Reuters

U.S. GDP grew at a 2.0% annualized rate in Q1, with AI-related business investment contributing 1.3 percentage points to the total growth. - Yahoo

US public debt hits 100.2% of GDP as Big Tech firms commit $690 billion to AI, competing for capital in the same debt markets. - TNW | Next-Featured

Google integrates Gemini AI into Workspace as privacy advocates warn that complex data retention policies and interface designs hinder user opt-out efforts. - Ars Technica

Apple expands its hardware roadmap with AI home hubs and smart glasses while refuting rumors that the Vision Pro line is being discontinued. - AppleInsider

Apple CEO Tim Cook expects Mac Studio and Mac mini supply shortages to last months following high demand for AI tools in Q2 2026. - MacRumors

Apple seems to be using Claude Code or possibly testing it for its support chat app, included .md files in a build. - Xcancel

finally caved and got a 2020 Mac Mini for OpenClaw experiments. seems like actually useful stuff I could do, message it to research stuff, update my Obsidian when something occurs to me, home automation, have automated shopping skills and whatnot.

Ray-banned: Meta contractor fired 1,100 AI trainers after being fired for revelation that Ray-Ban smart glasses recorded private and intimate footage. - TechSpot

Chinese courts ruled that replacing employees with AI is illegal, classifying automation as a controllable business strategy rather than an unavoidable disruption. - Caixin Global

AWS CEO Matt Garman says Amazon will hire 11,000 engineering interns in 2026, dismissing concerns that AI is eliminating software development roles. - Business Insider

AI sorting hat: McKinsey plans to use AI agents to help select staff for client teams. - Bloomberg

Gen Z’s AI frenemy: Students and workers report growing resentment toward AI tools over job displacement fears and social stigma despite high adoption rates. - The Verge

AI companies' pursuit of artificial general intelligence signals a tolerance for collateral damage, including the potential creation of a permanent underclass. - The New York Times

Senator Bernie Sanders called for an international AI treaty during a panel with Chinese scientists to address risks of mass unemployment and unregulated technology. - the Guardian

That’s Stochastic Parrot, Esq. to you, pal: Microsoft launched a Legal Agent for Word to automate contract reviews and risk assessments for US Frontier program members. - The Verge

It’s giving liability your honor: Vibe litigation democratizes justice. - Financial Times

will need more courts and judges, maybe Jevons applies, or maybe agentic arbitration.

Prompts for peace: New AI tools designed for deliberation aim to reduce political polarization and promote consensus. - Financial Times

News organizations push back against the use of a web archive for artificial intelligence training. - Bloomberg

Spotify is launching a green "Verified" badge to distinguish human artists from AI-generated content based on listener engagement and real-world presence. - the Guardian

GUARD Act requiring AI age checks advances out of committee - Electronic Frontier Foundation

Follow the latest AI headlines via druce.ai on Bluesky