AI hunts Firefox flaws, new enterprise AI hub & more

Enterprise AI is entering a new phase of capability and integration, as frontier models like OpenAI’s GPT-5.4 deliver unprecedented reasoning, coding and native computer interaction while Anthropic’s Claude Marketplace streamlines procurement of AI‑powered tools for large organizations. Simultaneously, advances in autonomous agent platforms, security‑focused AI applications, and hardware accelerators such as Apple’s M5 Max are expanding the operational reach of generative AI across business and research.

• OpenAI unveiled GPT-5.4, a frontier model that supports up to 1 M tokens, native desktop interaction and a tool‑search feature that cuts token usage by 47%, raising the benchmark ceiling for enterprise workloads.
• Anthropic launched the Claude Marketplace, centralizing procurement of Claude‑powered solutions from partners such as Snowflake, Replit and GitLab and simplifying billing for large organizations.
• Anthropic’s Claude identified 22 security flaws in Firefox, demonstrating AI‑driven vulnerability discovery as a new capability for cybersecurity teams.
• Lyzr secured a funding round valuing it at $250 million, underscoring growing investor confidence in autonomous AI agents for enterprise automation.
• Apple’s upcoming M5 Max chip delivered substantial CPU and on‑device AI inference gains, positioning the company as a contender in consumer‑grade AI hardware.
• Andrej Karpathy released autoresearch, an open‑source toolkit that lets AI agents autonomously train large language models on a single GPU, lowering research barriers.