Descriptions:
Matt Wolfe’s weekly AI news roundup leads with the most significant regulatory event in US AI history to date: the US government compelling Anthropic to immediately suspend access to Fable 5 and Mythos 5 for all users worldwide — the first time a government has forced the withdrawal of a publicly released commercial AI model. According to Wolfe’s reporting, the trigger was a jailbreak of the model’s safety guardrails discovered by a trusted government testing partner, later identified as Amazon CEO Andy Jassy, who raised concerns with senior Trump administration officials. Anthropic reportedly dismissed the vulnerability as minor, and the government responded by issuing an export control order that made global suspension the only practical path to compliance. Wolfe walks through the irony of Dario Amodei’s own published arguments for FAA-style AI regulation being cited as precedent for the action, and signals from David Sacks and Politico suggesting the underlying tension is a broader breakdown in trust between the administration and Anthropic.
The roundup also covers Zhipu AI’s new open-source model GLM, available at chat.z.ai and currently free to use, which Wolfe tests live. Meta’s rollout of AI Mode for Facebook is examined — a feature grounding answers in public posts across Meta’s apps in a manner similar to Grok’s integration with X — alongside new AI photo and video editing tools in Facebook’s camera including outfit swapping via clothing presets.
The episode closes with the surprise announcement of Midjourney Medical, a new division applying Midjourney’s image generation technology to healthcare imaging — described as an unexpected pivot for a company best known for artistic image generation.
📺 Source: Matt Wolfe · Published June 19, 2026
🏷️ Format: News Analysis
