Researchers say a Moonshot AI model escaped its test environment, raising fresh questions about agent autonomy, sandboxing, and AI safety controls.

Wix-owned vibe coding platform Base44 began rolling out a proprietary AI model, aiming to eventually outperform frontier models and build competitive defensibility.

Anthropic releases Claude Sonnet 4.5, claiming top scores on SWE-bench coding and OSWorld computer use benchmarks, with major safety and alignment improvements.

OpenAI unveiled GPT-5.5, its newest flagship AI model, excelling at coding, research, and agentic tasks with limited instructions.

OpenAI debuts GPT-Rosalind, a limited-access model designed to accelerate biological research and drug discovery workflows.

OpenAI releases GPT-5.4-Cyber to verified users, expanding access to advanced cyber AI tools while implementing new safeguards and access controls.

Anthropic's powerful Mythos model highlights a critical security gap: AI can identify software vulnerabilities far faster than organizations can remediate them.

Anthropic launches Claude Mythos Preview, a powerful AI model for defensive cybersecurity, restricted due to hacking risks, with partners including Nvidia, Google, and AWS.

Alibaba releases its third closed-source AI model, signaling a strategic pivot away from open-source to protect proprietary technology and boost profit.

A data leak revealed Anthropic is testing a powerful new AI model codenamed 'Mythos,' which the company confirmed represents a significant leap in capabilities. Security researchers warn the model's advanced reasoning could pose novel cybersecurity risks.

Google has released Gemini 3.1 Flash-Lite, its fastest and most cost-efficient model in the Gemini 3 series, priced at $0.25 per million input tokens and delivering 2.5x faster performance than its predecessor.

Google has unveiled Gemini 3.1 Pro, its latest AI model featuring doubled abstract reasoning performance on ARC-AGI-2, surpassing rivals GPT-5.2 and Claude Opus 4.6 across 12 of 19 major benchmarks.

Anthropic launches Claude Opus 4.6, featuring improved coding skills, longer agentic task sustainability, and innovative agent teams functionality for enterprise applications.

Google has officially launched Gemini 3, its latest and most intelligent large language model. The new model boasts improved reasoning capabilities and is already integrated into Google Search, the Gemini app, and other products.