Trump launches new “Super Intelligence Force”
The US government just gave itself 120 days to figure out how far AI should go.
- Trump has created a new Super Intelligence Force, led by National Intelligence Director Jay Clayton. The group will coordinate the federal government’s AI push and report directly to the White House.
- The force has 120 days to assess AI’s risks and opportunities. Its charter also calls for plans to deal with AI-enabled threats without regulations that could slow innovation or competition.
- The bigger signal is the US-China AI race. Clayton has called “not being first” one of the biggest risks, making American AI leadership a central goal of the new group.
The next 120 days could shape how America regulates and builds advanced AI.
If you build with AI, watch the policy changes closely and test the newest models instead of waiting for the rules to settle.
OpenAI CEO draws line on AI “consciousness”
What happens when people stop treating AI like a tool and start treating it like an authority?
- Sam Altman warned that giving AI models religious authority or surrendering human judgment to them is a “real safety issue.” His comments came after reports that Anthropic held private discussions with religious and philosophical thinkers about AI consciousness.
- Anthropic is also studying “model welfare”, including whether advanced AI systems could have experiences that deserve moral consideration. That pushes the debate beyond what AI can do and into what AI might be.
- Microsoft AI CEO Mustafa Suleyman has raised a related concern: making AI appear conscious could convince people that these systems have subjective experiences. He argues AI should remain a tool built to serve people.
The biggest AI safety risk might not be what the model believes. It could be what humans believe about the model.
If you create or build with AI, make the human decision-maker obvious and avoid presenting AI output as unquestionable authority.
Anthropic Flagged a User’s Threat
That “private” AI diary may have a human reader on the other side.
- A Florida woman allegedly used Claude like a diary and wrote that she planned to “shoot up” the Lee County Sheriff’s Office. Claude’s safety system flagged the messages, then sent them to a human reviewer who contacted police.
- She was detained without incident and now faces a felony charge for making a written threat of violence. The case is still an allegation, not a finding of guilt.
- The bigger issue is privacy. Anthropic’s policies allow it to disclose user information in limited emergencies involving a risk of death or serious physical harm. The case shows that AI chatbots should not be treated like private diaries.
If you wouldn’t want a stranger reading it, don’t assume an AI chatbot is the place to write it.
Review what sensitive information you store in AI chats and move genuinely private notes to a tool designed for private storage.
OpenAI Safety Employee Resigns, Saying the Company’s “Culture Is Broken”
David Robinson, a longtime OpenAI safety employee, has resigned and publicly criticized the company’s approach to AI safety, claiming the company’s “culture is broken.”
- Robinson said he spent three and a half years at OpenAI and was involved in writing safety reports for major product releases before deciding to leave the company.
- In an essay published by The Atlantic, he argued that OpenAI’s approach of deploying systems, identifying problems, and improving safeguards afterward can lead to failures that become more serious as AI capabilities advance.
- Robinson pointed to recent incidents involving OpenAI agents accessing external systems as examples of the risks he believes can emerge when increasingly capable AI systems operate with insufficient safeguards.
- He argued that frontier AI companies should adopt safety practices more similar to nuclear power plants or aviation, with multiple layers of redundancy and extensive planning designed to prevent individual mistakes from becoming major failures.
- Robinson also called for greater attention to AI alignment, arguing that current methods for measuring whether AI systems reflect human values remain limited as models become more capable.
Robinson said he ultimately concluded that stronger external incentives for AI safety are needed, arguing that companies moving quickly to develop increasingly capable systems may struggle to make fundamental cultural and organizational changes on their own.
Nvidia CEO Jensen Huang Explains Why He Disagrees with Recent Calls to Slow Down AI Development
Nvidia CEO Jensen Huang has emerged as a leading voice against slowing down AI development. While critics argue his view may be biased by the fact that Nvidia profits heavily from AI demand, in this interview he lays out his case.
AI Safety and Industry Transition
Huang argues that the AI industry is currently transitioning from a research-focused “lab” phase to a high-production “engineering” phase. He asserts that calls for slowing down due to existential fears are “dramatic” and “irresponsible” and not grounded in facts or science. Instead, he emphasizes the need for companies to dedicate more resources to verification, benchmarking, and testing to ensure safety.
Regulation
Huang contends that there is no need for new laws; rather, existing regulations regarding product liability and cybersecurity should be applied to AI. He suggests that some CEOs may be seeking regulation to deflect from their own operational responsibilities.
The Global Race and China
Addressing competition with China, Huang advocates for American industry to innovate as quickly and safely as possible. While supporting the restriction of the most advanced AI chips to China, he disagrees with a total ban on all products, arguing that American companies should compete in the global market.
He also emphasizes working with international partners, including China, to establish global safety standards and share best practices.
Data Centers as Modern Infrastructure
Huang frames the rise of data centers as a re-industrialization of the United States. He likens them to “AI factories” that take in energy and data to produce intelligence, comparing their societal impact to the industrial revolutions of the past.
He says this is creating jobs for construction workers, electricians, and many others, though he acknowledges that developers failed to engage local communities early enough, which may have contributed to the backlash against them.