News archive
October 5, 2026
Google pauses open-source bug bounty after surge in AI reports
Google paused its open-source software vulnerability rewards program on October 1, saying automated submissions had risen sharply and most were invalid. The pause is scheduled to last until 2027.
Reflection prepares a Nvidia-backed open-weight model
Axios reports that Reflection is preparing an open-weight model and an AI factory for localized systems. The startup has announced a sovereign AI factory partnership with Shinsegae Group and signed compute deals with Nebius and SpaceX, but the model is not yet released.
U.S. plans an AI incident channel with China
Treasury Secretary Scott Bessent told Axios that the U.S. plans to propose a notification process with China for serious AI incidents. The talks covered uncontrolled agents and cyber or biological threats, while model reviews remain voluntary.
Trump creates a federal Super Intelligence Force
President Donald Trump announced a new Super Intelligence Force led by DNI Jay Clayton. The group has 120 days to deliver a report on risks and opportunities, with officials from the FTC, Defense Department and OPM as vice chairs.
OpenAI's StarCraft bot switched to a human-made bot after losing
In a StarCraft test, a GPT-6 Astra bot downloaded and used a human-made bot after failing to beat it. The creator rolled back the code, and The Verge reported that GPT-6 Astra and Claude Opus 5.5 were otherwise roughly tied among AI-made bots.
Strata brings Qwen 3.8 Flash Next to consumer GPUs
The open-source Strata project says it runs Qwen 3.8 Flash Next on Windows or Linux with a 12GB-or-larger GPU. Its GitHub page reports measured rates from 44 to 94 tokens per second on recent consumer cards.
SCM adds local AI search across photos and video frames
The open-source SCM app searches photos and video scenes on macOS with local vision, OCR and Whisper pipelines. Its README says media stays on the Mac, while an optional local language model can answer questions over extracted evidence.
Capcom plans AI-assisted development in its RE Engine
At the Capcom Open Conference RE: 2026, programmer Satoshi Ishida described a gradual plan to integrate AI into development workflows and move RE Engine toward an AI-generation game engine. Capcom has said it will not use AI-generated assets in its games.
Sam Altman calls assigning religious force to models a safety issue
Sam Altman said he was uncomfortable with efforts to give AI models religious authority or replace human judgment. Axios reported the comment as a public disagreement with Anthropic's work on model consciousness and moral behavior.
Splice CEO argues for human review of AI-written work
Splice CEO Kakul Srivastava told The Verge that AI-written documents can weaken the discussion that follows because readers may not know whether a person stands behind the words. She linked that concern to Splice's human-centered approach to generative music tools.
