AI Trends: 6 October 2026

ChatGPT gets ads in image generation

OpenAI says it will test visual advertisements during ChatGPT image generation later in October, starting with a select group of US advertisers. The ads will be clearly labeled and kept separate from generated images. The company says ChatGPT now has 1.2 billion weekly users, which explains the push to monetize the free tier more aggressively.

Lab executives face New York City Council

Executives from Anthropic, OpenAI, Google and Meta testified at a New York City Council hearing on AI on 5 October, a sign that local governments are now pressing the major labs directly rather than waiting for federal rules.

Washington builds its AI command structure

The White House named Director of National Intelligence Jay Clayton as AI czar to lead a "Super Intelligence Force." Leadership is shared with FTC chair Andrew Ferguson, Pentagon research and engineering chief Emil Michael and OPM director Scott Kupor. Separately, California's SB 1246 sets fines of up to $10,000 per robotaxi that blocks emergency vehicles for more than 30 minutes, taking effect in July 2028.

Security: AI finds flaws, and agents become the risk

A vulnerability in Rejetto HFS was found with Anthropic's Mythos tool and exploited within a single day, showing how short the gap between discovery and attack has become. GitLab patched a critical CVSS 9.9 sandbox escape in its self-hosted AI gateway. Apple tightened macOS Full Disk Access permissions, citing the risk from AI agents.

Open models and the China gap

Nvidia-backed Reflection AI is preparing its first open-weight model, positioned against DeepSeek and Qwen for Western enterprises that want cost-effective, compliant bases. DeepSeek V4.1 Flash, released 10 September, scored 81.1 on LiveBench against Anthropic's 83.4. That is roughly a 3% gap, down from 9% in May and 15% earlier this year.

Business and legal

Roughly 30 "transfer stations" operating from a single office building in Beijing's Haidian district resell Claude access at 70 to 90% discounts through overseas accounts, and some reportedly pass off Qwen output as Claude. Former Groq engineers have sued the company's board in Delaware, alleging a $20 billion "non-exclusive" Nvidia deal moved nearly all engineers while short-changing shareholders. The DoJ opened a probe in September.

Research and tooling

Anthropic cofounder Christopher Olah reportedly held months of confidential meetings with Vatican theologians on AI consciousness, ahead of a May encyclical stating that AI systems "do not undergo experiences." On the tooling side, the open-source Strata engine runs the 125B-parameter Qwen3.8-Flash-Next mixture-of-experts model on a single 12GB consumer GPU with 32GB of system RAM by offloading experts, reaching about 100 to 140 tokens per second on an RTX 3090.

Compiled from public reporting for the 48 hours ending 6 October 2026.