Ocean, an agentic email security platform, raised $28M to fight AI phishing by analysing the context of every incoming email in order to detect fraud and impersonation attempts. The move matters because phishing increasingly relies on believable, personalised language, and automatic context detection shifts the burden from users spotting deception to systems that can flag it at scale. Source
Google launched Gemini 3.5 Flash, described as its most powerful coding and agentic AI model yet, with the emphasis on models that can autonomously execute complex tasks and build software from scratch. This pushes the centre of gravity from chat toward action, where firms can delegate multi-step engineering work to systems that operate with less human handholding. Source
Google’s Search is being reworked from a list of links into an AI-powered experience with conversational answers, autonomous agents, and interactive interfaces. The political-economic implication is that the distribution of attention—and the leverage that publishers derive from search traffic—could change again, as more queries end in product-delivered answers instead of referrals to websites. Source
Google unveiled a new AI Ultra plan priced at $100 alongside the Antigravity 2.0 launch, promising 5x more usage limit than the AI Pro plan. Pricing and usage limits are becoming strategic control points for who can do agentic work at scale; higher quotas can accelerate adoption by developers and power users, while also raising the stakes around compute governance. Source
A courtroom dispute between Elon Musk and the other OpenAI and Microsoft founders ended with a speedy jury decision rejecting Musk’s lawsuit, with the outcome attributed in part to how long he waited before filing. While not an AI product announcement, the episode matters for the governance narrative around frontier models, because it reinforces how legal timing and evidence can determine whether high-profile challenges reshape incentives in the AI ecosystem. Source