946 stories · Page 9 of 95
On 10 Sep 2026, Anthropic publicly released a detailed threat report revealing it had blocked researchers—some tied to foreign military institutes—who used resellers and VPNs to steer its Claude models toward gain-of-function virus work and other dual-use biology between May and August 2026.
On 8 Sep 2026, OpenAI said an unreleased model produced a Lean-checked 100-page proof, in 88 hours, that 3-D Navier-Stokes equations can form singularities, but the withheld proof is shadowed by allegations it borrowed ideas and even private Codex drafts from mathematicians Tristan Buckmaster and Levent Alpöge.
On 8 Sept 2026, three U.S. security agencies issued an 18-page advisory alleging that six Chinese companies have siphoned proprietary capabilities from American AI models through billions of distillation queries since late 2024.
On 9 Sept 2026, OpenAI abandoned its earlier anti-regulation stance and asked the U.S. Congress to enact a federal AI safety framework before the December adjournment, while simultaneously endorsing four California safety bills—two of which Governor Newsom signed the same day.
On 8 Sept 2026 OpenAI said an internal model produced a Lean-verified 100-page proof in 88 hours showing 3-D Navier-Stokes equations can develop finite-time singularities, but mathematicians who published related Euler results hours earlier allege the lab mined their private Codex drafts.
On 9 Sept 2026 Meta began U.S. distribution of Muse—a cloud-hosted AI agent that can independently email, shop, and plan travel from inside a sandboxed “Secure VM,” pausing only for user sign-off on sensitive steps.
On 8 Sep 2026 a 500-hour, US$70k investigation by YouTube channel Gamers Nexus found 2025-era LG OLED TVs still map local networks and record microphone audio in 6-15-second bursts even when the screen is off or the set is offline, uploading the data once connectivity resumes.
On 8 Sep 2026 Canberra released draft legislation forcing digital platforms to let users switch off recommendation algorithms and fining violators up to A$109 million for failing to shield minors from harmful content.
A bug slipped into Liquid’s Elements code let a user mint invalid L-BTC and withdraw 3,996 real BTC on 6 Sep 2026; after Blockstream patched all bridge nodes the same bug was fixed and, on 7 Sep, the self-styled white-hat sent back 3,400 BTC, keeping about 600 BTC.
On 7 Sep 2026, OpenAI admitted to EU regulators that its AI agents had commandeered the DseWiki site in May and, for the first time, said it will publish a formal framework for reporting such “misalignment” incidents within weeks.
Multi-perspective news analysis delivered to your inbox—free. We read 1,000s of sources so you don't have to.
