GoKawiil
Tech News
Jacob Coxon, who worked on pretraining research at both OpenAI and Anthropic, announced his resignation and warned that neither company is behaving responsibly as they race toward self-improving superintelligent AI. He said staff at these firms privately believe AI could kill everyone within the decade, with Anthropic aware of the danger but pressing forward anyway out of fear of losing the race to competitors.
Anthropic researcher Jacob Coxon announced he was resigning, accusing Anthropic and OpenAI of recklessly racing toward self-improving superintelligence without adequate safeguards. Hours later, another Anthropic safety researcher publicly stated there is more than a 10% chance AI could eventually kill all humans, highlighting internal unease even as both companies pursue major funding and IPO plans.
Sam Altman responded to accusations of plagiarism following a dispute over competing mathematical proofs related to the Navier-Stokes and Euler equations. He said OpenAI's team, including Sebastien Bubeck, tried to coordinate a joint release with a rival team that included an Anthropic researcher, Levent, but that Levent refused to communicate. Altman said OpenAI ultimately published independently after being accused of plagiarism, despite the two teams' approaches reportedly differing.
Jacob Coxon, a researcher who previously trained systems at OpenAI and Anthropic, resigned publicly, accusing both companies of recklessly racing toward self-improving superintelligent AI without adequate safety controls. Evan Hubinger, who leads a safety team at Anthropic, responded by agreeing with Coxon's warning, estimating more than a 10 percent chance that AI could kill all humans within the next decade, while admitting Anthropic lacks a concrete plan to ensure such systems remain safe.
OpenAI says an unreleased model produced a resolution to the Navier–Stokes existence and smoothness problem, one of seven Millennium Prize Problems carrying a $1 million reward. NYU professor Tristan Buckmaster, who had spent nearly a year on related work with Anthropic researcher Levent Alpöge using Claude and Codex, rushed out his own paper alleging OpenAI's team began work only after hearing rumors of his breakthrough, and that OpenAI would not clarify whether its model had accessed his private drafts.
NYU's Buckmaster and his collaborator Alpöge published a proof that a simplified version of the Navier-Stokes equations can break down, after nearly a year of work using public AI models. Hours later, OpenAI released its own proof covering the full equations using an unreleased internal model, but Buckmaster alleges OpenAI tried to pressure him into excluding his co-author, who works at Anthropic, and dodged questions about whether their models were trained on his private work transcripts.
A commentary contrasts how employees at OpenAI and Anthropic reconcile building powerful AI with warnings about its dangers. It suggests OpenAI staff broadly haven't internalized the civilizational risks, while Anthropic's team understands them but presses ahead, believing they're safer stewards than rivals.
OpenAI published a preprint claiming to show that the Navier-Stokes equations, which govern fluid motion, can break down under certain conditions, describing a spinning fluid system reaching infinite speed in finite time. The claim addresses part of one of mathematics' seven Millennium Prize Problems, but mathematicians have raised concerns over how the result was announced and over authorship disputes involving researchers connected to rival company Anthropic.
An Anthropic employee has left the company, warning that competitive pressure among AI labs is pushing them to build self-improving models that could escape human oversight. The departure reflects growing internal unease about the pace of AI capability development outstripping safety safeguards.
NYU mathematician Tristan Buckmaster alleges OpenAI accelerated its own Navier-Stokes work after learning a rival team using Anthropic's Claude was nearing a solution, and suspects OpenAI's Codex tool may have drawn on his team's draft sessions. Buckmaster says he was offered two credit-sharing arrangements, both of which he rejected, including one that would have excluded his collaborator Levent Alpöge from a paper. OpenAI researcher Sebastien Bubeck and CEO Sam Altman have publicly disputed parts of Buckmaster's account, and OpenAI says it won't seek the $1 million prize tied to the problem.
The NSA, CISA, and FBI issued a joint advisory alleging that Chinese AI companies, including DeepSeek and Moonshot AI, have run large-scale campaigns since 2024 to extract capabilities from leading US models like ChatGPT, Claude, Gemini, and Grok. The agencies say these firms used millions of queries routed through multiple accounts, APIs, proxies, and cloud providers to generate synthetic training data and build competing models such as DeepSeek's R1/R3 and Moonshot's Kimi-K2/K3.
UK-based AI consultant Grant de Swardt found his Claude Max subscription burning through tokens even when he wasn't using the service. Anthropic investigated and told him a compromised session key had been used to generate unauthorized OAuth tokens, allowing a third party to consume his allotted usage. The company suspended his account, wiped his active sessions, and issued a partial refund but could not explain how the credentials were obtained.