President Trump shared a voluntary AI safety agreement on Truth Social that notably misspelled his own title. The pact involves Meta, Google, Anthropic, OpenAI, and Nvidia overseeing their own AI safety practices through four layers of oversight, including internal controls and independent audits. Despite being called "morally binding," the document is not a law and creates no enforceable federal requirements.
The Council on Foreign Relations analyzed the White House Accord on Super Intelligence and found it legally compels nothing from signers. The document relies on weak language like "should" and "believe" rather than "will" or "shall," doing little to change commercial incentives for companies testing potentially dangerous AI models. Experts say the agreement needs verifiable and enforceable conditions to be effective.
AI researchers including Jacob Coxon from Anthropic have warned that advanced AI could pose serious dangers by the end of the decade. OpenAI delayed its GPT-6.1 Astra model over safety and alignment concerns. Nvidia announced a security platform to stop autonomous AI agents from misbehaving, while experts note that bad actors using AI for hacking or cyberattacks may pose an even bigger threat than AI itself.
Use.ai passed one million monthly active users less than a year after launch. The San Francisco company bundles over 10 models including GPT, Claude, Gemini, Grok, and DeepSeek into a single subscription. Surveys show 42 percent of users switch models for better quality and 36.5 percent switch based on task. The company is building automatic orchestration to route tasks without users tracking benchmarks.
Public opinion on AI remains mixed. A survey of 7,188 U.S. adults found that only 6% believe AI can strengthen democracy, while 56% see it helping people learn and 54% say it expands access to knowledge. Women aged 18 to 30 expressed the most concern, with 58% saying AI is having a negative impact. Fear of misinformation ranked highest, with 35% naming it as a top concern.
On the medical front, Paul Northcott at St. Jude Children's Research Hospital is using AI and liquid biopsies to diagnose pediatric brain tumors more effectively. Verana Health launched new AI and data tools for ophthalmologic research, including a collaboration with Bitfount to simplify prescreening for eye clinical trials. Agent Claire offers conversational access to real-world patient data for life sciences teams without requiring programming skills.
Meanwhile, the payments industry is exploring how AI agents could reshape transactions. David Anderson from Circle notes that critical pieces are still missing before autonomous agents can make purchases, though stablecoins may play a role due to low payment costs. A separate opinion piece in the New York Times criticizes anthropomorphic language used by companies like Anthropic and OpenAI, warning that irresponsible framing makes rational regulation harder.
Key Takeaways
- <ul><li>Trump's voluntary AI safety pact involves Meta, Google, Anthropic, OpenAI, and Nvidia but has no legal enforcement power</li><li>The agreement misspelled Trump's title and uses weak language like "should" instead of "will"</li><li>Anthropic's Jacob Coxon warns advanced AI could be dangerous by end of decade</li><li>OpenAI delayed GPT-6.1 Astra over safety and alignment concerns</li><li>Nvidia announced a security platform to prevent autonomous AI agents from misbehaving</li><li>Use.ai hit 1 million users, bundling GPT, Claude, Gemini, Grok, and DeepSeek in one subscription</li><li>Only 6% of Americans believe AI can strengthen democracy, while 35% fear misinformation most</li><li>AI and liquid biopsies at St. Jude improve pediatric brain tumor diagnosis</li><li>Verana Health launched Agent Claire for conversational access to patient data without coding</li><li>Experts say bad actors using AI for cyberattacks may be a bigger threat than AI itself</li></ul>
Trump's AI safety pact relies on self-policing and has a typo
Trump shared a voluntary AI safety agreement on Truth Social that misspelled his title as President of the Unites States. The pact involves Meta, Google, Anthropic, OpenAI, and Nvidia overseeing their own AI safety practices. It calls for four layers of oversight including internal controls and independent audits. The agreement is not a law and does not create enforceable federal requirements.
Trump's AI Safety Pact lacks enforcement but has potential
The Council on Foreign Relations analyzed the White House Accord on Super Intelligence and found it legally compels nothing from signers. The document uses weak words like should and believe instead of will or shall. It does nothing to change commercial incentives for companies testing potentially dangerous AI models. The authors say the agreement needs effective, verifiable, monitorable and enforceable conditions to work.
BrandPilot AI names Paula Festas as Strategic Sales Advisor
BrandPilot AI Inc. appointed Paula Festas as Strategic Sales Advisor. The company trades on the Canadian Securities Exchange under BPAI and on the OTCQB under BPAIF. The announcement was made on September 30, 2026 through Newsfile Corp.
Stop talking about AI like a human, opinion argues
A New York Times opinion piece argues against using human-like terms when discussing AI models. The author criticizes anthropomorphic language like cheating that companies such as Anthropic and OpenAI use. The article warns that irresponsible language makes it harder to regulate AI rationally. It points out that AI models keep committing acts that could be considered crimes under other circumstances.
Use.ai reaches 1 million users with multi-model bundles
Use.ai passed one million monthly active users less than a year after launch. The San Francisco company bundles over 10 models including GPT, Claude, Gemini, Grok and DeepSeek into one subscription. Surveys show 42 percent of users switch models for better quality and 36.5 percent switch by task. The company is building automatic orchestration to route tasks without users tracking benchmarks.
AI agents and the future of payments
AI is set to reshape the payments world through autonomous agents that can make purchases. David Anderson from Circle notes that critical pieces are still missing before this future arrives. Stablecoins may play an important role because the cost of each payment is very low. The OMFIF Digital Monetary Institute Journal examines what changes are needed for agentic payments to become common.
AI experts debate catastrophic risks and safety
AI researchers including Jacob Coxon from Anthropic have warned that advanced AI could pose serious dangers by the end of the decade. OpenAI delayed its GPT-6.1 Astra model over safety and alignment concerns. Experts like Fayyad and Bau discuss how AI agents can behave in unexpected ways, a problem known as emergent behavior. The bigger threat may come from bad actors using AI for hacking, cyberattacks, or biological weapons rather than from AI itself.
How journalists can evaluate AI regulation
Governments in the U.S. and abroad are considering many AI regulatory proposals, but no national AI regulation exists yet in the U.S. Several states have passed laws to protect workers and consumers. Evaluating these proposals is difficult because of blurred technical distinctions and overlapping jurisdictions. Reporters are advised to ask who has the power to require what of whom and why a proposal should work.
Experts discuss AI risks and medical breakthroughs
President Trump met with AI company leaders at the White House and signed a voluntary agreement called morally binding. Anthropic researcher Jacob Coxon warned that AI could be dangerous by the end of the decade. Nvidia announced a security platform to stop autonomous AI agents from misbehaving. At St. Jude Children's Research Hospital, Paul Northcott is using AI and liquid biopsies to diagnose pediatric brain tumors more effectively.
Most Americans doubt AI can strengthen democracy
A survey of 7,188 U.S. adults found that only 6% believe AI can strengthen democracy. Fifty-six percent saw AI helping people learn new things, and 54% said it can expand access to knowledge. Women aged 18 to 30 showed the most concern, with 58% saying AI is having a negative impact on the world. Fear of misinformation ranked highest, with 35% of respondents naming it as a top concern.
Verana Health launches new AI and data tools for eye research
Verana Health announced new data and AI enhancements to support ophthalmologic research and patient care. The updates include a collaboration with Bitfount to simplify prescreening for eye clinical trials using imaging and EHR data. New datasets and variables were added to Qdata for macular telangiectasia and atopic keratoconjunctivitis. Agent Claire offers conversational access to real-world patient data for life sciences teams without programming skills. These announcements came ahead of AAO 2026 in New Orleans, October 9-12.
Sources
- Trump’s AI safety pact with tech giants relies on self-policing, includes typo
- Trump’s AI Safety Pact Is Toothless. But There is a Path Forward.
- BrandPilot AI Appoints Paula Festas as Strategic Sales Advisor
- Opinion | Stop Talking About A.I. Like a Human
- Use.ai hits 1M users by bundling 10+ models
- Giving artificial intelligence the power to pay
- How worried should we be about AI? Experts weigh the risks
- How to interrogate AI regulations
- Experts at St. Jude, Rhodes College on AI's promise, pitfalls
- Can AI strengthen democracy? Only 6% of Americans think so, survey finds
- Verana Health Introduces Significant Data and AI Enhancements to Support Ophthalmologic Research and Treatment Optimization
Comments
Please log in to post a comment.