India's Sarvam AI hit unicorn status, OpenAI launched a cybersecurity arm, Colorado scrapped its own AI law, and the EU's enforcement deadline is now six weeks away.
Sarvam AI is India's newest AI unicorn, valued at $1.5 billion
Bengaluru-based Sarvam AI closed the first tranche of its $300M Series B at a $1.5 billion valuation, with HCL Tech, Bessemer Venture Partners, and Khosla Ventures among the backers. Sarvam builds open large language models tuned for Indian languages and designed to run on lower-cost edge hardware, making it the centrepiece of India's sovereign AI ambitions.
Why it matters: India now has a homegrown AI company valued at over a billion dollars whose entire mission is building AI that works in Indian languages, for Indian contexts, on Indian infrastructure. That matters for anyone who has ever felt that global AI tools don't quite fit.
OpenAI launches Daybreak, its first dedicated cybersecurity program
OpenAI announced Daybreak, a cybersecurity initiative combining GPT-5.5 models with Codex security tools to automate threat detection, vulnerability patching, and remediation workflows for enterprise clients. The company also revealed OpenAI DeployCo, a majority-owned consulting subsidiary backed by over $4 billion, designed to help large organisations actually implement AI at scale.
Why it matters: OpenAI is no longer just a model-builder, it's becoming a full-service AI company. The cybersecurity move signals that AI is now being embedded into the systems that protect critical digital infrastructure worldwide.
Colorado scrapped its own AI law and replaced it with something lighter
Colorado's landmark AI Act, which was due to take effect on June 30, 2026, was repealed and replaced with SB 26-189 before it could ever be enforced. The new law strips out complex risk-management requirements and replaces them with simpler consumer notice obligations, you must be told when AI is being used to make decisions about you, and you can ask for a human to review the outcome.
Why it matters: It shows just how hard it is to turn AI regulation into law, even in states that tried early. The softened version is a signal that business lobbying is winning the first round, but consumer rights groups are watching closely.
The EU AI Act becomes fully enforceable in six weeks, here's what that means
The European Union's AI Act will be fully applicable from 2 August 2026, making it the world's most comprehensive binding AI law. Companies operating in Europe must now demonstrate compliance with risk classifications, transparency requirements, and prohibited AI uses, including social scoring and real-time biometric surveillance in public spaces.
Why it matters: Any company, anywhere in the world, that sells AI products or services to European users is now on the clock. Non-compliance can carry fines of up to €35 million or 7% of global revenue, whichever is higher.
US senators probe Big Tech over soaring electricity bills tied to AI data centres
Three Democratic senators launched an investigation into the role of Big Tech data centres in driving up electricity costs for households and businesses. AI data centres already consume over 4% of all US electricity, government models suggest this could triple to 12% by 2028. Some individual AI video generation tasks consume energy equivalent to charging a smartphone hundreds of times.
Why it matters: The energy cost of AI is no longer an abstract concern. It's affecting electricity grids and household bills, and regulators are starting to connect the dots between AI infrastructure expansion and the public interest.
IndiaAI Mission now targets 100,000 public GPUs by December 2026
India's IndiaAI Mission has scaled to 54,000 GPUs and has set a target of 100,000 public GPUs by December 2026. Alongside the compute push, the government is running workforce programs including SOAR (Skilling for AI Readiness), YUVAi for younger learners, and AI fellowships supporting 500 PhD scholars, 5,000 postgraduates, and 8,000 undergraduates. AI labs are being set up in Tier-2 and Tier-3 cities.
Why it matters: India is trying to build AI infrastructure and AI talent at the same time. The Tier-2 and Tier-3 city focus matters, it signals that the government wants AI skills to reach beyond the major metros.
Anthropic is quietly heading for a near-trillion-dollar IPO
Anthropic has confidentially filed a draft registration statement for an IPO, following its $65B Series H that valued the company at $965 billion. If it lists publicly, Anthropic would become one of the most valuable companies ever to go public. The IPO would mark the moment AI companies stop being private experiments and become part of the public stock market, with all the scrutiny, quarterly pressure, and retail investor access that comes with it.
NYC schools require AI bias audits + Claude Mythos 1 quietly launches
New York City's Department of Education has mandated that all AI tools used in its 1.1-million-student system must pass a bias and equity review before deployment, a compliance playbook is coming later this month. Separately, Anthropic released Claude Mythos 1, a new model in its June 2026 wave, with a focus on reasoning and long-context tasks. Details on availability and pricing are still thin.