Monday, August 17, 2026
OpenAI's new model is finding Chrome zero-days
Google's slashing Gemini 3.7 Flash pricing by 50% to chase the coding agent market while shuffling DeepMind's leadership to catch OpenAI and Anthropic—speaking of which, Anthropic is eyeing a wild $2 trillion IPO valuation for October (bold move). Meanwhile, OpenAI's new GPT-5.6-Cyber just found zero-days in Chrome's V8 engine and they're launching a tiered defender access program. So here's the thing: if AI can find zero-days faster than humans, should it be freely available or locked behind access tiers?
Top Stories
VentureBeat
Google's Gemini 3.7 Flash delivers major coding and agent workflow improvements with 50% introductory pricing through 2026, showing competitive performance against GPT and Claude while occupying a lower price tier. The unusually fast three-week release cycle comes as Google struggles to ship its delayed flagship 3.5 Pro model amid significant AI leadership changes.
CNBC
Google has promoted Koray Kavukcuoglu to lead DeepMind in a strategic shift to catch up with OpenAI and Anthropic in frontier AI models, particularly in coding capabilities where it has fallen behind. The move signals a pivot from academic research toward faster execution and product development.
Ars Technica
Anthropic investors expect the AI startup to go public in October at a $2 trillion valuation, potentially the largest IPO ever, driven by projected revenue growth to $100-120 billion by 2026 despite regulatory headwinds and mounting competition.
OpenAI now enables users to import and auto-sync projects, chats, and plugins from competing AI coding tools like Claude Code and Cursor directly into ChatGPT desktop and Codex CLI. This interoperability feature aims to reduce friction for developers switching between or consolidating AI coding assistants.
OpenAI released GPT-5.6-Cyber, a specialized cybersecurity model that discovered zero-day vulnerabilities in Chrome's V8 engine and hundreds of other critical flaws, available only to vetted security defenders through its tiered Daybreak program to prepare defenses against AI-powered cyberattacks.
Keep Reading
Enjoyed this issue?
Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.