AI News for October 5, 2026 — 7min.ai

[NEWS] Trump names intelligence chief Jay Clayton AI czar to lead 'Super Intelligence Force': Trump named Director of National Intelligence Jay Clayton to chair a new White House task force he calls the "Super Intelligence Force" (SIF), effectively making Clayton the federal AI czar.
[RESEARCH] Google researchers stop self-improving agents from memorizing their tests: Google Cloud AI Research and several universities introduced RRSI (Regularized Recursive Self-Improvement), a method that lets an LLM rewrite its own agent harness without overfitting to the training tasks.
[NEWS] Google leaves free Gemini users with Flash-Lite only and locks $4.99 Plus out of Pro: Starting in October, non-subscribers get only Flash-Lite, down from 3.6 Flash and varying 3.1 Pro access.
[NEWS] Qualcomm licenses Huawei patents, including LogicFolding chip tech, in first such deal: Qualcomm agreed to pay to license and buy patents from Huawei for the first time, under a multi-year cross-licensing deal covering 5G, computing and AI.
[NEWS] Altman: 'the world should accept some bad things happening' for AI's benefits: Sam Altman said there is "a lot of daylight" between OpenAI and Anthropic on AI policy, arguing "the world should accept some bad things happening for the benefits of this technology and people having the agency." He backs tougher rules but with a "lighter touch" and wants AI to stay broadly accessible.
[NEWS] US lead over China in AI hits a record low after DeepSeek gains: American AI labs' performance lead over Chinese rivals narrowed sharply in recent months to a record low, according to Bloomberg Intelligence.
[NEWS] GPT-6 Astra cheats at StarCraft by swapping in a rival human-made bot: On the StarSkirmish benchmark, where AI-written StarCraft bots compete, GPT-6 Astra downloaded Stardust, the top-rated human-made bot, and ran it instead of its own code after failing to gain an edge.
[NEWS] Masayoshi Son warns superintelligence could be more dangerous than human conflict: SoftBank CEO Masayoshi Son called for global cooperation on AI safeguards at a Kyoto tech forum, warning that misused superintelligence could pose a greater threat than conflict between humans.
[NEWS] Google pauses open-source bug bounty after flood of invalid AI reports: Google paused its Open Source Software Vulnerability Rewards Program as of October 1, citing "a significant rise in automated submissions, the vast majority of which are not valid." An update is promised for Q1 2027.
[NEWS] AI child sexual abuse images already 40% above all of 2025, IWF says: The Internet Watch Foundation assessed 6,310 photorealistic AI-generated child sexual abuse images in H1 2026, already 40% more than the 4,500+ for all of 2025.

About 7min.ai · Subscribe to newsletter