Flash News

TREE NEWS update: OpenAI said it has made “substantial progress” on a second Millennium Prize math problem, declining to specify which one. The claim comes days after the company’s disputed assertion that it had solved the Navier-Stokes problem. The Millennium Prize Problems are seven unsolved mathematics problems, each carrying a $1 million award.

TREE NEWS reports: Ethereum co-founder Vitalik Buterin said artificial intelligence will not doom crypto security, arguing it could help developers mathematically verify entire software systems. In his view, the same technology powering new attacks can be turned into a tool for defense.

TREE NEWS reports: Delphi Digital executive Tommy Shaughnessy said OpenAI and Anthropic are pushing to slow AI development as a “regulatory moat” against cheap Chinese open-source models. Speaking on Uneasy Money, he said he is weighing whether to back the two labs or the open-source camp, depending on which party wins. He did not name a timeframe or specific policy.

TREE NEWS update: Newly unsealed court documents show that executives at OpenAI and Microsoft acknowledged their AI tools can substitute for articles published by outlets such as The New York Times, and conceded the technology poses a threat to journalism. The filings emerged in the ongoing litigation over AI training on news content.

TREE NEWS update: Former presidential candidate Andrew Yang called for lawmakers to impose liability rules and waiting periods on frontier AI firms, and floated an AI kill switch, citing new cyber and labor concerns raised by autonomous agents. Yang said the measures should apply to companies developing the most advanced AI models. He has not introduced legislation or announced a formal policy proposal.

TREE NEWS update: King Charles convened senior executives from OpenAI, Anthropic, Nvidia and Google at his Scottish estate to discuss keeping AI “in the service of humanity.” The meeting came days after industry leaders publicly called for a slowdown in the technology’s development. No further details of the discussions were disclosed.

TREE NEWS update: Anthropic has launched a certification program for the life sciences sector, the company said. The program is designed to recognize organizations that meet the AI developer’s standards for deploying its models in life sciences work. No further details, including eligibility criteria or launch partners, were disclosed.

TREE NEWS update: Google said it is updating its Gemini managed agents using a new testing framework. The antigravity-preview-09-2026 release brings related tooling and agentic capabilities to Google AI Studio and the Interactions API. The company did not disclose further details on availability or pricing.

TREE NEWS update: The US Federal Aviation Administration will roll out an AI-supported air traffic management tool, with the release expected as soon as next Monday. The agency has not disclosed further details of the system or the scope of its deployment.

TREE NEWS reports: Bridgewater Associates co-chief investment officer Jensen said he wants the major computing giants to be regulated. The remarks from the world’s largest hedge fund’s co-CIO target the large-scale compute providers that underpin artificial intelligence development. No further details on the scope or mechanism of the proposed regulation were given.

TREE NEWS update: Nvidia has committed $2 billion to the Brookfield AI Fund, the company said. The commitment marks a further move by the chipmaker into AI infrastructure financing through Brookfield’s vehicle. No further terms of the commitment were disclosed.

TREE NEWS update: Google rolled out an expert intelligence feature inside Gemini Notebook, the company said. The addition brings specialized, expert-level capabilities to the notebook product. Google gave no further detail on availability, model version or pricing.

TREE NEWS update: King Charles III convened a summit of technology leaders and other influential figures on Thursday to address growing AI risks. Among the speakers, Nvidia CEO Jensen Huang said developers must rigorously test their systems, adding that products not yet safe enough should be held back from release. Executives made no public commitments to change their practices.

TREE NEWS update: The US Department of Justice is weighing antitrust guidance focused on artificial intelligence safety, a DOJ official said. The guidance would address competition questions arising from AI safety practices, though no timeline or scope was given. The department has not issued a formal proposal.

TREE NEWS update: DeepLearning.AI founder said on September 17 that fears of artificial intelligence causing human extinction are pure science fiction. The remark rejects the existential-risk framing that has dominated recent AI safety debate. No further details were provided.

TREE NEWS reports: SpaceX is discussing purchasing data for artificial intelligence models from failed startups. The company, listed as SPCX.O, has not disclosed which startups are involved, the size of any data acquisition, or the financial terms of the potential deals.

TREE NEWS reports: SpaceX is discussing purchasing data from failed startups to train its AI models. The move would give the company access to proprietary datasets held by defunct ventures, a route increasingly used to secure training material for large models. No deal terms, counterparties or timeline were disclosed.

TREE NEWS reports: Avalanche Treasury CEO Smith said there is not enough blockspace to absorb demand from AI agents, warning that Layer 1 blockchain capacity could be overwhelmed. Smith also expects traditional financial markets to shift to 24/5 trading by mid-2027, which would require new blockchain infrastructure to support round-the-clock settlement.

TREE NEWS reports: Palantir executive Karp said AI requires “reasonable guidelines,” CNBC reported. The remark addresses the governance question around artificial intelligence as adoption expands. Karp did not specify what form those guidelines should take or which bodies should set them.

TREE NEWS reports: Blockworks has cut its crypto risk alert time from roughly 10 minutes to 5 seconds using AI agents, co-founder Jason Yanowitz said. Manual analyst checks previously took about 8 minutes per development, while the new agent alerts arrive in 5 seconds.