Google's Gemini 3.5 Flash set new benchmarks, Anthropic confidentially filed for a near-$1 trillion IPO, and US law now lets deepfake victims sue their harassers.
Gemini 3.5 Flash is here, and it's outperforming models twice its size
Google released Gemini 3.5 Flash, a model built for speed that still rivals much larger flagship AI systems. It beat Gemini 3.1 Pro on key coding and reasoning tests. A more powerful Gemini 3.5 Pro is on its way later this month, alongside Gemini Omni, a multimodal model that can take in and create video, audio, text, and images at once. Developers can already access Managed Agents in the Gemini API, which provision a full cloud environment for AI to plan, code, and browse the web autonomously.
Why it matters: Google is closing the gap between fast, cheap models and frontier-level performance. For anyone building AI tools or using AI at work, this means capable AI is getting more affordable and accessible, not less.
Anthropic files for IPO at nearly $1 trillion. OpenAI expected next.
Anthropic, the company behind Claude, filed confidentially to go public on June 1. The company is currently valued at close to $965 billion and grew its annual revenue from $9 billion to over $30 billion in just a few months. OpenAI (valued at $852 billion) is widely expected to follow. If both go public, they would each be worth more than ExxonMobil or JPMorgan Chase.
Why it matters: The two most important AI companies in the world are about to become publicly traded. That means more transparency, more public pressure to perform, and more scrutiny of safety and ethics claims from shareholders and regulators alike.
Colorado's AI Act kicks in June 30, the first state law with real teeth
Colorado's AI Act becomes enforceable on June 30, 2026. Companies using AI in high-stakes decisions (hiring, loans, healthcare) must now run formal risk assessments, document how their systems work, and prove they are not discriminating based on protected characteristics. The law was delayed from its original February start date after businesses pushed back, but it is now final.
Why it matters: This is the first comprehensive state-level AI law in the US with clear obligations and penalties. Other states (and global regulators) are watching closely. If your company uses AI to make decisions about people, this is the shape of things to come everywhere.
The DEFIANCE Act is law: deepfake victims can now sue in federal court
The US Senate passed the DEFIANCE Act unanimously earlier this year, and it is now in effect. It creates a federal right for people to sue anyone who creates, shares, or hosts non-consensual sexually explicit deepfakes of them. Damages can reach $150,000, or $250,000 if the deepfake was used alongside stalking, harassment, or assault. A separate AI bill advancing through Congress would also crack down on deepfake distribution more broadly and strengthen whistleblower protections for AI workers.
Why it matters: For years, victims of AI-generated abuse had almost no legal recourse. That has now changed in the US. India and EU legislators are watching this framework to model their own responses.
Deepfake job candidates are getting through interviews, and companies can't tell
Hiring fraud using AI deepfakes is surging in 2026. AI tools now let fake candidates generate perfectly tailored resumes and swap their face and voice in real-time during video interviews. Once hired, these fraudulent employees gain access to internal systems. Fraud analysts now expect this to be the biggest AI-driven hiring risk of the year, with no easy technical fix yet available to most employers.
Why it matters: If you are in HR, recruiting, or run a remote team, this is no longer a future risk. Verifying candidates in person or with identity-checked video calls is becoming a necessary step, not an optional one.
Rajasthan launches its own AI policy. India pushes for homegrown foundation models.
The Rajasthan AI-ML Policy 2026 was unveiled at a regional AI conference in Jaipur, covering responsible AI use in public services, employment, and skill development. Separately, India's Office of the Principal Scientific Adviser released a white paper calling for India to build its own AI foundation models (trained on India-specific data) to improve transparency, inclusivity, and national strategic independence. The India AI Mission has a budget of over ₹10,000 crore to support this.
Why it matters: India is moving from consuming global AI tools to building infrastructure on its own terms. State-level AI policies mean that what AI can and cannot do in India will increasingly be shaped by Indian priorities, not Silicon Valley defaults.
Two near-$1 trillion AI companies going public. What happens next?
Anthropic's confidential IPO filing sets up one of the most anticipated public listings in years. OpenAI is expected to follow. Together, these two companies would represent well over $1.8 trillion in market value entering public markets, likely by autumn. Watch for what happens to safety commitments, open-source access, and pricing when quarterly earnings reports and shareholder pressure become part of the picture. The AI industry's shift from private to public is the slow-moving story that will reshape everything.
GPT-5.3-Codex drops + Google trains 6M teachers in AI + Grok 4.20 arrives
OpenAI launched GPT-5.3-Codex, combining its best reasoning and coding stacks into one model that runs about 25% faster than its predecessor, aimed squarely at developers and engineers. Google's free AI literacy training programme for all 6 million K-12 and higher education teachers in the US went live in May and is now fully available. And xAI released Grok 4.20 with multi-agent capabilities, continuing Elon Musk's effort to keep pace with OpenAI and Google in the model race.