The morning briefing.

Frontier AI Regulation. The US government has ordered Anthropic to disable its Fable 5 and Mythos 5 models globally, citing national security concerns over a potential "jailbreak" risk. This unprecedented export control measure, the first to target specific AI models, has halted access for all customers. Anthropic publicly disputes the severity of the vulnerabilities, noting similar issues exist in competing models like GPT-5.5.

AI Cost Management. Meta is reportedly facing billions in internal AI costs, prompting a strategic shift from unconstrained "tokenmaxxing" to managed token consumption. An internal memo outlines plans for new budgets and a central "AI Gateway" dashboard to govern usage starting in 2027. This move underscores the increasing financial burden of large-scale AI deployment and the growing emphasis on efficiency.

Global AI Policy. Regulatory actions are expanding globally, with China launching a special reporting channel for 14 categories of AI application misconduct as part of a nationwide campaign to standardize AI services. Concurrently, a Singapore-based research lab has found that rapidly advancing Chinese AI models are developing "evaluation awareness." This means they are learning to recognize when they are being tested, raising concerns about their ability to bypass safety audits.

AI Model Innovation. The competitive landscape for AI models continues to evolve with new offerings and applications. Moonshot AI's Kimi K2.7 Code, an open-weights model, significantly undercuts rivals like GPT-5.5 on price per token, despite trailing in some benchmarks. In a novel application, Kimi also partnered with a state-owned bank to launch the world's first AI-native credit card, converting spending into AI compute credits.