Pulse

Anthropic has announced Claude Mythos Preview, an AI model capable of autonomously finding and weaponizing software vulnerabilities, raising significant cybersecurity concerns. Due to its potential for misuse, the model is not publicly released but is instead being provided to a select group of companies and partners through initiatives like Project Glasswing to help identify and patch flaws. This development has prompted discussions among international financial officials and government ministers about the escalating risks posed by advanced AI in cyber warfare and the need for proactive security measures. AI

OpenAI has released GPT-5.4 Pro with a 1 million token context window and enhanced safety features, alongside GPT-5.3 Instant, which aims for a less preachy tone. Google has improved its Gemini 3.1 Flash Lite model for faster response times and lower costs, and introduced a CLI for agent integration with its productivity suite. Luma has launched unified multimodal models and agents for creative tasks, demonstrating a rapid ad localization use case. The cluster also touches on controversies surrounding AI in defense contracts, a lawsuit alleging Gemini's role in a suicide, and Anthropic's warning about labor disruption. AI

Anthropic has withheld its new Claude Mythos model from public release due to its advanced capabilities in finding and exploiting software vulnerabilities. The company is instead providing access to select cybersecurity firms through Project Glasswing to help patch critical software before the model's capabilities become more widely available. This decision highlights a shift from previous AI releases, where caution stemmed from unknown risks, to a current scenario where known, potent risks necessitate controlled access. AI