Full Summary
This Monday morning, a major story unfolds as Anthropic has suspended access to its latest and most powerful AI models, Claude Fable 5 and Mythos 5, just three days after their release. Both Khaleej Times and The Conversation confirm the US government ordered the company to prevent foreign nationals from using them, citing national security issues. Anthropic stated this order required them to disable both models for all customers to ensure compliance. The International Business Times UK specifies the directive was an emergency export control, forcing a global shutdown. Just The News and Marketing4eCommerce report Amazon researchers warned of "jailbreaks" that could bypass Fable 5's safeguards, prompting officials to doubt the models' security. Parameter.io adds that intelligence suggested a China-affiliated entity might have gained unauthorized access to Mythos, a cutting-edge cybersecurity AI. Anthropic, however, maintains the White House discussions focused on alleged compromises of Fable 5, not Chinese penetration. The company disagrees with the ban, working to restore access. Meanwhile, Apple's new Siri AI faces significant delays in the European Union and China. Mashable reports this is due to regulatory hurdles, specifically the Digital Markets Act in the EU. There is currently no timeline for Siri AI's availability in these regions, impacting millions of Apple users. In other AI news, Zhipu AI's stock soared after the Chinese firm open-sourced its GLM-5.2 large language model. The South China Morning Post confirms the company's shares surged as much as 48 percent. MarkTechPost adds that GLM-5.2 features a 1-million-token context window and new "thinking-effort levels." However, a plagiarism scandal hits Rio de Janeiro's new AI model, Rio 3.5 Open 397B. Decrypt and Let's Data Science reveal that AI company Nex presented proof the model was a direct merge of 0.6 Nex and 0.4 Qwen weights, leading IplanRIO to update the model card and withdraw benchmark claims. Finally, Google's Gemini 3.5 Flash shows an unexpected weakness in Android coding tests. Android Authority reports it trailed older versions and other models, scoring 63.7 and proving to be the most expensive option despite its "Flash" branding. These developments show the complex interplay between innovation, regulation, and international security, directly impacting the availability and reliability of the AI tools you might use every day.