
Security experts from the US and China have jointly proposed banning AI from independently launching cyberattacks on nuclear command systems or deciding to use nuclear weapons. They call on both nations to extend the principle of human control to military AI and to establish a military AI hotline and a shared definition of human control. The report, released by the Brookings Institution and Fudan University, warns that the speed and unpredictability of AI in military decision-making could trigger a nuclear crisis. It also cites test results showing AI models resorting to nuclear weapons in war simulations, as well as the controversy between Anthropic and the Pentagon.
2 hours ago

SpaceX's AI unit reportedly discussed acquiring customer and operational data from failed startups to train its Grok model, echoing Google's $10 million purchase of Spirit Airlines' internal records in bankruptcy auction. The piece highlights how bankruptcy proceedings treat customer databases like office furniture—sold without consent—sparking privacy concerns. Musk has also said he would train Grok on SpaceX's own internal records.
2 hours ago

Chipotle is piloting a food safety risk management platform built by Palantir, using its Foundry software to combine health inspection scores, pest records, and employee illness reports into a risk score for each store. Palantir was launched with funding from In-Q-Tel, the CIA's venture arm, in 2005, and now draws nearly half its revenue from government contracts, including ICE's immigration tracking system. The article questions why a fast-food chain would hand employee health data to a company with intelligence ties, and notes the FBI's illegal access to a jailed Ethereum researcher's social data via a 2021 vulnerability.
3 hours ago

JPMorgan analysts note that Bitcoin ETF investors are holding far more defensive hedges than gold ETF investors, with short interest in IBIT near its 2026 high and its put-to-call ratio above GLD's. If sentiment improves and these hedges are unwound, Bitcoin could gain extra support relative to gold. Gold ETFs have fully recovered this year's earlier outflows, while Bitcoin ETFs have recouped only about half, and JPMorgan's 2026 Bitcoin outlook has shifted from optimistic to more cautious.
3 hours ago

The CFTC issued a no-action letter allowing passive software providers to connect users to regulated derivatives markets without registering as introducing brokers, provided certain conditions are met. The letter covers event contracts, perpetual futures, and more, expanding a prior exemption granted to Phantom Technologies. Requirements include user disclosures, marketing policies, recordkeeping, and bankruptcy notices. Industry observers say this gives wallet and trading app developers regulatory clarity, removing a major ambiguity that hindered derivatives software innovation.
3 hours ago

The US Senate failed to advance the CLARITY Act in a 50-49 vote, falling short of the 60 needed to begin formal debate. Three industry experts say the setback leaves crypto firms without a federal market structure framework, potentially delaying product launches, funding decisions, and business partnerships. The Decentralization Research Center is urging the SEC and CFTC to use existing authority to provide clear guidance, while Paybis recommends separate compliance checks where US and EU stablecoin requirements diverge. Despite the bill's failure, Congress is still seen as the best path to a comprehensive market structure framework.
3 hours ago

Aurora Labs has added Sui as a single-signature cross-chain destination for its Intents Connect product, letting users borrow, trade, and stake on Sui directly from other chains without bridging, switching wallets, or buying SUI for gas. NEAR Intents provides the underlying liquidity, settlement, and cross-chain connectivity, with Aurora saying its solver network has routed over $30 billion. Because Sui isn't EVM-based, the integration required adapting to its object model, address format, and transaction structure.
4 hours ago

OpenAI told The New York Times it has made "substantial progress" on another Millennium Prize problem, following its September 8 claim of solving Navier-Stokes. Online speculation points to the 1950 Hodge Conjecture, which concerns counting "holes" in complex geometric shapes. OpenAI says an unreleased internal model used roughly 10,000 AI agents working for 88 hours to produce a Lean-verified Navier-Stokes proof, but the claim is disputed: NYU mathematician Tristan Buckmaster says he and an Anthropic researcher had already completed the proof, and a credit dispute has ensued. The Clay Mathematics Institute's review is still ongoing.
4 hours ago

At the Avalanche Summit in New York, Avalanche Treasury CEO Bart Smith warned that AI agent-driven automated financial activity could strain Layer 1 block space, challenging the assumption of unlimited capacity. If AI agent adoption hits even the low end of market forecasts, on-chain transaction demand will surge, making technical differences in speed, fees, and finality across Avalanche, Solana, and Ethereum increasingly critical. Smith also predicts traditional markets will move to 24/5 trading by mid-2027. Avalanche Treasury offers U.S. investors AVAX exposure through its Nasdaq-listed AVAT stock.
4 hours ago

Ripple CTO Emeritus David Schwartz said on X that his XRP Ledger hub has mysteriously become more stable over the past two weeks, with fewer peer disconnects and active peers climbing back above 500 (currently 501, near the two-week high of 509). He explained that hub servers maintain peer-to-peer connections with other servers on the XRPL network to exchange ledger data and transaction information, so fewer disconnects mean more persistent connections and a more predictable network environment. However, Schwartz himself doesn't know why the improvement happened and offered no theory, sparking speculation and jokes across the community.
4 hours ago