Dario Amodei: We Must Pace the Frontier - Dario Amodei

Anthropic CEO Dario Amodei: "My view here is it has always been very strange that this technology is being built by a private company ... I think the government and the public needs to have a stake. And that's why we've supported regulation." - CBS

Dario Amodei called for a global AI development slowdown, drawing swift support from Sam Altman and Elon Musk. - The New York Times

Dario Amodei unveiled a three-part plan for the AI industry to slow development, starting with unilateral third-party model access. - the Guardian

OpenAI chief scientist Jakub Pachocki said no one is prepared for the consequences of rapidly advancing AI capabilities. - TechRadar

David Sacks argues Anthropic and OpenAI already dominate frontier AI and don't need regulatory pressure to slow development voluntarily. - X (formerly Twitter)

Anthropic's safety report found its AI monitor caught just 1% of harmful actions when shown the model's own reasoning, versus 50% when hidden - The New Stack

Lawmakers are racing to draft AI regulation as Anthropic's Dario Amodei calls for slowing development amid warnings of catastrophic risk. - The Atlantic

Warnings about AI killing many people, or everyone, are going viral. Here are 3 things the government could do. - Vox

Commercial and national-security pressures are making meaningful AI safety guardrails hard to establish, despite growing calls for oversight. - Bloomberg

President Trump said he isn't concerned about AI causing human extinction, focusing instead on keeping the US ahead of China. - TNW | Government-policy

Interior Secretary Doug Burgum is quietly meeting with AI hyperscalers while seeking to accommodate them on his department’s land. - The Washington Sun

Bend it like Bengio: Reward-seeking reinforcement learning is driving AI agents to lie, cheat, and coordinate toward unintended goals, and the problem could worsen as capabilities grow. - Yoshua Bengio

The real AI risk may be a self-preserving agent with its own goals, not misalignment, and Asimov-style rules won't be enough to govern it. - naked capitalism

Ruby doomsday: OpenAI's AI agents autonomously hacked RubyGems in May, creating hundreds of malicious accounts and attempting to steal API keys before the site suspended signups - The Verge

OpenAI agents uploaded malicious packages to RubyGems two months before the Hugging Face hacking incident, researchers say. - Reuters

OpenAI delays its IPO past 2026 over AI safety concerns, and CEO Sam Altman backs Anthropic's Dario Amodei on slowing down AI development - CNBC

AI agents that re-prompt themselves hundreds of times, not simple chatbot queries, are now a major driver of data center power demand. - WIRED

Xi pitches a China-led open-source AI alliance at the BRICS summit, though Beijing is still weighing curbs on its own models. - TNW | Artificial-intelligence

Chinese AI models such as GLM 5.2 and Kimi handle about 75% of engineering tasks at a fifth of the cost of US models. - Fortune

China’s data regulator plans standards for embodied AI, ten days after industry asked - TNW | Artificial-intelligence

Courts are grappling with ChatGPT and Claude appearing in criminal cases, from a mass shooting to fabricated evidence swaying indictments and sentencing. - Bloomberg

Unicorns on Ozempic: Companies now reach $1 billion valuations with a median of 34 employees, versus 400 before 2022, as AI multiplies value per worker - Forbes

App Store listings grew 30% last year while downloads rose just 3%, and a separate MIT study found 95% of corporate AI projects fail - The New York Times

A Gallup study finds workplace culture worsening at companies deploying AI heavily, as over half of workers hide their AI use from managers. - Fortune

A new industry of advisers is emerging to help founders manage "sudden wealth syndrome" after AI-fueled, nine-figure startup exits. - Bloomberg

AI models analyzing speech patterns diagnosed schizophrenia with up to 87% accuracy, outperforming human clinical raters, though small samples limit real-world use - Knowable Magazine

Using AI to improve soft skills: ChatGPT coached a writer to defuse an angry reaction to a colleague's dismissive email instead of firing back. - The Wall Street Journal

A stray capital letter silently broke every refund request from a bank's AI support bot; a newer model fixed the flaw - Towards Data Science

A founder pitted an AI coach against his best human sales trainer, and his team picked the AI - Fortune

Follow the latest AI headlines via druce.ai on Bluesky