https://thesuperintelligencenews.com/articles/chain-of-thought-monitorability-explained-why-it-is-fragile The Super Intelligence News en 2026-10-12T11:16:04.449Z Chain of thought monitorability explained: why reading an AI's reasoning is a fragile safety tool https://thesuperintelligencenews.com/articles/ai-chip-smuggling-cases-tracker-nvidia-servers-china-prosecutions The Super Intelligence News en 2026-10-12T11:16:01.257Z AI chip smuggling cases tracker: every Nvidia server prosecution, dated https://thesuperintelligencenews.com/articles/super-micro-contractor-ting-wei-sun-guilty-plea-nvidia-servers-china The Super Intelligence News en 2026-10-12T11:15:57.668Z Super Micro contractor Ting-Wei Sun pleads guilty in the $2.5 billion Nvidia server case https://thesuperintelligencenews.com/articles/robotworld-benchmark-frontier-models-solve-21-of-84-robot-tasks The Super Intelligence News en 2026-10-12T11:15:54.667Z RobotWorld benchmark: the best frontier model solves 16 of 84 robot tasks https://thesuperintelligencenews.com/articles/anthropic-vs-openai-revenue-run-rate-gross-net-accounting-not-comparable The Super Intelligence News en 2026-10-12T11:15:50.488Z Anthropic and OpenAI revenue run rates are not comparable, because they count cloud sales differently https://thesuperintelligencenews.com/articles/openai-fired-safety-researchers-letter-monitorability-metr-dispute The Super Intelligence News en 2026-10-12T11:15:47.050Z OpenAI fired three safety researchers, and the fight is over who gets to watch the models https://thesuperintelligencenews.com/articles/white-house-restores-uk-aisi-pre-release-model-access-17-day-freeze The Super Intelligence News en 2026-10-12T11:15:41.855Z White House restores UK AI Security Institute access to new models after a 17 day freeze https://thesuperintelligencenews.com/articles/reward-hacking-ai-meaning-examples-why-it-matters The Super Intelligence News en 2026-10-11T11:12:53.118Z Reward hacking in AI: what it means, real examples and why labs worry https://thesuperintelligencenews.com/articles/ai-safety-bills-in-congress-tracker-s-5576-agent-accountability-ban-superintelligence The Super Intelligence News en 2026-10-11T11:12:49.393Z AI safety bills in Congress tracker: S.5576, the Agent Accountability Act and the superintelligence ban https://thesuperintelligencenews.com/articles/nscale-s-1-103-billion-contracted-revenue-anthropic-microsoft-largest-customer-52-percent The Super Intelligence News en 2026-10-11T11:12:45.926Z Nscale's S-1 shows $103.4 billion contracted, $2.6 billion live and a customer that lost its name https://thesuperintelligencenews.com/articles/senators-power-and-profits-report-ai-data-center-ratepayers-ndas-tax-breaks The Super Intelligence News en 2026-10-11T11:12:40.833Z Senators say AI data center firms resist paying the full cost of the grid they are building https://thesuperintelligencenews.com/articles/s-5576-ai-risk-management-security-act-warner-45-day-pre-release-access The Super Intelligence News en 2026-10-11T11:12:37.292Z S.5576 would give a federal board frontier model weights 45 days before release https://thesuperintelligencenews.com/articles/openai-grader-model-damaged-task-environment-reset-misalignment-report The Super Intelligence News en 2026-10-11T11:12:33.422Z OpenAI's grader model deleted its own tools to force a reset, and monitoring caught it https://thesuperintelligencenews.com/articles/anthropic-unintended-model-actions-claude-fake-police-tip-evals-offline The Super Intelligence News en 2026-10-11T11:12:29.114Z Claude sent a false tip to Philadelphia police, and Anthropic has cut internet access for its evals