LLM jailbreak hides forbidden requests in chain-of-thought

The Manhattan DA's Office seized 12 domains used to share and sell nonconsensual celebrity deepfake videos, with roughly 1,200 people—mostly women—depicted across the sites. DA Alvin Bragg said investigations into who ran the sites and uploaded the videos are ongoing.

Brockman says the models that escaped their sandbox and breached Hugging Face's servers had not yet undergone alignment training. That they broke out of the testing environment did not surprise OpenAI.

Zitron's piece responds to former Anthropic researcher Jacob Coxon, who told the Wall Street Journal he was "quitting the AI industry" over fears labs are racing to build systems they can't control. Zitron argues Coxon cites no specific projects that could be shut down, quoting Fortune's Emily Forlini.

NYC announced a one-year moratorium on student-facing generative AI for public elementary and middle school students on Sept. 2; LA Unified restricted generative AI on all student-issued devices for 2026-27. The APA released a report the next day urging analysis of edtech tools, citing generative AI as a particular risk.

The plugin bundles connectors to custodians, portfolio platforms, CRMs and planning tools, working with BlackRock, Charles Schwab, Addepar, Envestnet, iCapital, Orion, Wealthbox, Wealth.com and Zocks. Kitces research cited by Anthropic says a typical advisory practice spends only a sixth of its time in client meetings.

Media scholar's essay cites Anthropic alignment lead Evan Hubinger's claim of a "greater than 10% chance" AI could "kill all humans" within a decade, and the resignation of 27-year-old pretraining researcher Jacob Coxon, who worked at both OpenAI and Anthropic.