AI Safety Splits Washington as Industry Debate Intensifies

The most consequential AI development in the evidence is political: President Donald Trump has turned safety warnings into a direct challenge to the industry’s calls for restraint, while Congress signals that some form of oversight may still be possible. At the same time, leading AI companies are discussing joint safety work and shipping products that make the policy argument more immediate—from Apple’s redesigned Siri to humanoid robots intended to work beside people.

Trump Rejects AI Safety Warnings as Congress Hesitates

Trump has described concerns about AI escaping human control and calls for stronger safeguards as a “hoax,” arguing that rapid development is necessary to preserve the United States’ lead over China. At the All-In Summit, he made the position unusually public by taking a call from Nvidia CEO Jensen Huang during a live discussion and dismissing AI dangers before a technology audience, as Axios reported.

The administration’s position is meeting a less categorical response in Congress. Senate Majority Leader John Thune said lawmakers have a role and that guardrails for frontier models could make sense, while House Speaker Mike Johnson has called for an industry meeting before legislation. The result is not a settled policy direction but a widening gap between the president’s rhetoric and lawmakers’ attempts to preserve room for oversight, according to Axios.

OpenAI, Anthropic, and Google Discuss Joint Safety Work

OpenAI, Anthropic, and Google DeepMind have been in talks for weeks about collaboration on AI safety issues, with discussions linked to a proposal from Google’s Demis Hassabis for a US-led standards body. CNBC reported that a spokesperson confirmed the discussions.

The talks matter because they show that the companies’ public warnings are not only positioning for regulation: at least some of the leading labs are also exploring shared safety practices. But the evidence describes discussions and joint efforts, not a completed standards institution or binding agreement. That distinction will be important as Washington weighs voluntary corporate safeguards against formal rules.

Nvidia Argues Market Forces Can Handle AI Safety

Jensen Huang is taking the opposing industry line. Nvidia’s CEO said AI does not need new laws and argued that safety can be engineered by the companies building each product, rather than imposed through additional regulation. Bloomberg reported that Huang believes market forces can support safe innovation.

Huang’s position gives the voluntary-safety argument one of the industry’s most prominent voices, but it also sharpens the conflict with the labs discussing shared safety work and with lawmakers who see frontier systems as requiring guardrails. The immediate question is whether company-by-company controls can address risks that cross products, firms, and national borders.

Apple Puts Its Redesigned Siri Into Beta

Apple has begun rolling out a redesigned Siri AI beta with daily limits, following the release of iOS 27 and ahead of the iPhone 18 launch. Users may need to join a waitlist, and reporting indicates Apple is considering a paid tier for heavier use. CNBC reported that the test is now reaching users.

The rollout makes the AI business model more tangible: Apple is testing how an assistant embedded in a mass-market operating system can be limited, expanded, and potentially monetized. The beta’s restrictions also underline that consumer deployment is arriving with practical capacity and product-design constraints, not as an unlimited capability.

Agility’s Digit 5 Is Designed to Work Beside People

Agility Robotics has unveiled Digit 5, a humanoid robot designed to operate near human coworkers without the physical cages used in many industrial deployments. According to Ars Technica, the robot can stop, move away, or squat and assume a seated position when a person gets too close.

That is a concrete shift in the deployment problem: the challenge is not only whether a robot can perform a task, but whether it can share a workspace safely. Business Insider reported that Agility says Digit robots are deployed at nine customer facilities in North America, while the company is trying to turn early deployments into a larger business. The safety behavior is therefore both a technical feature and a condition for expanding where humanoids can work.

Meta Packages More AI Access Into Paid Subscriptions

Meta has launched Meta One subscription bundles that combine expanded access to its AI tools with premium features across Facebook, Instagram, and WhatsApp. The plans target individual users, creators, businesses, and AI power users, while Meta says the core experience of its apps and Meta AI remains free. The Verge reported that the bundles are available globally and may eventually include Edits and AI glasses.

The move places AI usage inside a broader subscription strategy rather than treating the assistant as a standalone product. It also sets up a test of whether consumers and businesses will pay for higher limits and bundled access when basic AI functionality remains free.

The Next Test Is Implementation

Today’s evidence points to a specific tension rather than a settled AI-policy consensus. The White House is rejecting safety warnings in the language of political combat, Nvidia is arguing for company-led controls, and OpenAI, Anthropic, and Google are discussing cooperation without yet announcing a binding framework. Meanwhile, product rollouts are moving ahead in consumer software, subscriptions, and workplaces. The next thing to watch is whether voluntary safety practices become concrete and interoperable enough to satisfy lawmakers—or whether deployments and political pressure outpace that process.