OpenAI DevDay is live today with a dozen products expected, three AI giants are quietly building their own safety rulebook, and Anthropic shipped faster and cheaper models.
OpenAI DevDay is live today: an always-on AI agent rumoured
OpenAI's biggest developer event of the year is underway in San Francisco, with Sam Altman's keynote live-streamed at 10am PT. Leaks point to a dozen-plus product announcements including an always-on AI agent called "O" that runs continuously in the background, a new $500/month ChatGPT Pro Max tier, and expanded API access for developers.
Why it matters: Every year DevDay rewires what developers build. Whatever lands today will shape the AI tools students and professionals are using six months from now, so keep an eye on the announcements as they drop.
OpenAI, Anthropic, and Google are writing AI's rulebook, themselves
The three biggest AI labs confirmed weeks of behind-the-scenes talks and plan to launch a self-regulatory body called the Standards Authority for Frontier AI (SAFA) by end of year or early 2027. The initiative began when Google DeepMind CEO Demis Hassabis proposed a FINRA-style standards body in July. This week, Anthropic CEO Dario Amodei publicly called on labs to slow development of their most advanced models, an appeal endorsed by both Hassabis and Sam Altman.
Why it matters: This is the most significant coordination between competing AI companies yet. Self-regulation sounds bureaucratic, but it decides how fast AI moves and whether governments step in with harder rules next.
Claude Sonnet 5.5 is 30% faster and 30% cheaper, and a Marketplace just launched
Anthropic released Claude Sonnet 5.5, its latest mid-tier model, generating output more than 30% faster and costing up to 30% less per task than its predecessor. Claude Opus 5.5 also arrived at roughly 40% lower cost. Alongside the models, Anthropic launched the Claude Marketplace, a catalogue of tools and agents built on Claude for use across industries.
Why it matters: Better models at lower prices means the AI tools you already use (writing assistants, tutors, research helpers) get smarter without getting more expensive. The Marketplace is where that translates into ready-to-use applications.
You can now buy from Flipkart directly inside Google's AI search
Google is testing a feature in India where shoppers can purchase products from Flipkart without leaving Gemini or Google's AI Mode. A "Buy" button appears on Flipkart product listings inside the AI interface, routing users to a Flipkart checkout without switching apps. It is one of the first large-scale AI-native commerce tests in India.
Why it matters: This is AI moving from answering questions to completing transactions. India is one of the first markets where this is being tested at scale, which means Indian consumers and businesses will shape what it looks like globally.
California now requires risk checks before AI chatbots reach children
Governor Newsom signed legislation requiring tech companies to conduct safety risk assessments before deploying AI chatbots to young people, with fines up to $1 million per child for violations. It is one of the strictest child-AI protection laws passed by any US state and takes effect immediately.
Why it matters: California's tech laws often set a national template. If you are a parent, educator, or anyone building tools used in schools, the standard for deploying AI to minors just became much stricter, and more other states are watching.
Instinct raises $1 billion: betting the next interface is your personal AI agent
Instinct, a startup building personal AI agents for everyday life, raised $1 billion in a Series C round at a $10 billion valuation. Personal AI agents are software that runs continuously on your behalf (booking, tracking, reminding), without you needing to open an app or type a prompt. The raise is one of the largest in the fast-growing agentic AI category.
Why it matters: When investors write $1 billion cheques for "an AI that manages your life", the product is no longer a concept. Expect these tools in mainstream consumer hands within the next 12–18 months.
AI's biggest players are writing their own rules, will it stick?
SAFA (the self-regulatory body being built by OpenAI, Anthropic, and Google) is the most ambitious experiment in AI industry governance yet. Three companies that compete intensely, each with real financial incentives to move fast, trying to collectively agree on when to slow down. The test is whether the standards they set have teeth, or become a soft barrier designed to delay harder government regulation. The next three to six months will reveal which it is.
Google retires Gems, EU AI Act fully live, and September's model flood continues
Google is migrating Gemini Gems (its custom AI versions) to a new format called "Skills" from November 17. The EU AI Act reached its final compliance phase on August 2, meaning most AI deployments across Europe are now regulated. September has already seen 24 new AI models from 16 providers, with GPT-6 Sol, GPT-6 Luna, Grok 4.7, and DeepSeek V4.1-Flash among recent arrivals.