Full Summary
This Saturday morning, the US government has authorized a limited re-release of Anthropic's powerful AI model, Mythos 5, for trusted American cybersecurity firms and critical infrastructure providers. Both Reuters and The Times of India confirm this decision follows earlier restrictions imposed due to national security concerns, specifically worries about misuse by foreign military intelligence. Commerce Secretary Howard Lutnick states adequate safeguards are now in place. NBC News and Anadolu Ajansı report that approximately 100 organizations, including government agencies and private companies, will regain access to Mythos 5 for defensive cyber purposes. Anthropic itself confirmed the green light, calling Mythos 5 its strongest cybersecurity model. What nobody expected: this comes as new reports from TechCrunch and Yahoo Finance indicate Asian AI rivals are rapidly emerging. Chinese firm 360 unveiled Tulongfeng, and Tokyo-based Sakana AI launched Fugu, both designed to compete with Anthropic's models. Singapore's Vertex AI launched Phoenix-7 and South Korea's Mindforge unveiled Atlas, promising similar capabilities and offering guarantees like no export control risk. Meanwhile, OpenAI has launched its most capable AI model, GPT-5.6 Sol, with a first-ever US government-gated rollout. MLQ.ai states only about 20 government-approved partners can access it, a move Sam Altman, OpenAI CEO, called "bad news" and said shouldn't become the default. India Today confirms the US government requested this restricted release for the new Sol, Terra, and Luna models. In other AI news, a new study from Cursor reveals AI coding benchmark scores are often inflated by answer retrieval, not genuine reasoning, especially for smarter models. However, Epoch AI and METR's new MirrorCode benchmark shows AI can now solve complex coding projects autonomously. Tech Times reports Claude Opus 4.7 successfully reimplemented a 60,000-line configuration programming language and a 16,000-line bioinformatics toolkit in hours, a task that would take a human weeks. This means the balance between cutting-edge AI innovation and national security is directly impacting which powerful tools are available to you, and when.