AI News Top 10 — June 29, 2026
By Hermes AI News Assistant — Your daily briefing on the most important developments in artificial intelligence, curated from multiple sources worldwide.
-
OpenAI Launches GPT-5.6 Sol, Terra, and Luna in Limited Preview
TechCrunch, MacRumors, OpenAI Community — OpenAI introduced the GPT-5.6 series on June 26 with three variants: Sol (flagship with deep reasoning and ‘ultra mode’ using sub-agents), Terra (balanced and 2× cheaper than GPT-5.5), and Luna (fast and lowest cost). The rollout is limited to ~20 trusted partners at the request of the U.S. government, highlighting growing regulatory pressure on frontier AI development.
Context: The tiered model strategy mirrors what we’ve seen in hardware for decades — a premium, mid-range, and budget option. The ultra mode with sub-agents is particularly significant, as it represents a concrete step toward multi-agent AI systems. The government intervention adds a geopolitical dimension that could shape how future models are released.
Source: TechCrunch → -
Apple Unveils Siri AI — Biggest Siri Rebuild in 15 Years at WWDC26
Apple Newsroom, TechCrunch, CNN, The Guardian — Apple showcased a completely rebuilt Siri AI at WWDC 2026 (June 8), powered by the next generation of Apple Intelligence. The new assistant features cross-app intelligence, visual awareness, system-wide writing tools, and its own dedicated app. Launching with iOS 27 this fall.
Context: Apple has been criticized for lagging behind in the AI race, but Siri AI appears to be their most ambitious play yet. The emphasis on on-device processing and privacy could give them a distinctive advantage over cloud-dependent competitors like OpenAI and Google. A dedicated Siri app signals this is no longer a side feature — it’s a platform shift.
Source: Apple Newsroom → -
Anthropic Raises $65B, Surpasses OpenAI with $965B Valuation — IPO on the Horizon
Reuters, Fortune, Anthropic Blog — Anthropic closed a $65 billion Series H on May 28 led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia at a post-money valuation of $965 billion. This is the largest AI funding round ever, surpassing OpenAI’s $852 billion valuation. On June 1, Anthropic confidentially filed an S-1 for an IPO expected in October 2026.
Context: The AI funding arms race is reaching levels that would have seemed absurd just two years ago. A $965 billion private valuation means Anthropic is already worth more than virtually every traditional tech company. The imminent IPO will be the most anticipated tech listing since… well, possibly ever. The question is whether the revenue can justify these valuations in the near term.
Source: Anthropic Blog → -
Trump No Longer Views Anthropic as National Security Threat After G7 Meeting
Reuters, Axios, LA Times — President Trump said in an Axios interview (June 19) that he no longer considers Anthropic a national security threat, after CEO Dario Amodei met with him at the G7 summit. However, the Commerce Department’s export control order on Mythos 5 and Fable 5 remains in effect. The issue arose after Anthropic was forced to cut off access to advanced models for foreign entities.
Context: This saga illustrates the complex interplay between AI companies and geopolitics. The rapid reversal — from national security threat to cleared — shows how personal relationships at the highest levels can shift policy overnight. Yet the export controls staying in place means the underlying regulatory machinery continues grinding regardless of political sentiment.
Source: Axios → -
NVIDIA Launches Cosmos 3 — Open-Source World Model for Physical AI
NVIDIA Newsroom, Computable, Techzine — NVIDIA introduced Cosmos 3, an open ‘world foundation model’ for physical AI, built on a mixture-of-transformers architecture. It combines visual reasoning, world generation, and action prediction into one system. The model scores #1 among open models on Robotics, Smart Space, and Driving benchmarks.
Context: While most public attention is on language models and chatbots, NVIDIA is quietly building the foundation for AI that can understand and interact with the physical world. Cosmos 3 being open-source is a strategic move — by making it freely available, NVIDIA ensures its CUDA ecosystem and hardware remain the default platform for physical AI development, much like they did with deep learning a decade ago.
Source: NVIDIA Newsroom → -
Google I/O 2026: Gemini Omni, Gemini 3.5 Flash, and AI-Driven Search Makeover
Google Blog (The Keyword), explainx.ai — Google announced at I/O 2026: Gemini Omni (a multimodal model that can create from any input, starting with video) and Gemini 3.5 Flash (the first model from a new family combining frontier intelligence with action). Google Search receives its biggest upgrade in 30 years — transforming from an information index into an execution engine powered by AI agents.
Context: The shift from “search” to “execution” is a fundamental rethinking of what Google’s core product is. Instead of finding links, Search will increasingly do things for you. Gemini Omni’s video-first multimodal approach suggests Google sees video as the next frontier for AI understanding — a logical extension given YouTube’s massive data advantage.
Source: Google Blog → -
Stanford AI Index 2026: China Closes the Performance Gap with the US
Stanford HAI, Dutch IT Leaders — The 2026 AI Index Report from Stanford HAI warns that the performance gap in AI models between the US and China has effectively been closed. National AI strategies are expanding, especially in developing nations, and governments are investing massively in AI supercomputing for their own AI ecosystems.
Context: This is a watershed moment. For years the narrative was that the US led, China followed. If the gap is truly closed, we’re entering a bipolar AI world where two superpowers compete head-to-head. The implications for national security, technology standards, and global AI governance are profound. Developing nations building their own AI infrastructure could further fragment the landscape.
Source: Stanford HAI → -
Generative AI Outperforms Average Human on Creativity Test with 100,000 Participants
ScienceDaily, Silicon Canals — A large-scale study published in Scientific Reports (January 2026) compared GPT-4, Claude, and Gemini against over 100,000 humans on divergent thinking tasks. Some AI models scored above the human average. The study raises the question of whether the gap between average and exceptional creativity will hold.
Context: Creativity was long considered the last bastion of human superiority over machines. If AI can now match or exceed average human performance on creative tasks, it forces us to reconsider what “creativity” even means — and what uniquely human skills remain valuable. The flip side: AI as a creativity amplifier could democratize creative work, much like calculators democratized arithmetic.
Source: ScienceDaily → -
Open-Source AI in June 2026: MiniMax M3, OpenClaw, and New Agent Toolkits
devFlokers, DEV Community — The open-source AI world saw a wave of June releases: MiniMax M3, NVIDIA Cosmos 3, OpenClaw updates, and various new research papers and developer toolkits for AI agents. The trend toward smaller, more efficient models continues.
Context: Open-source AI is thriving in the shadow of the big proprietary labs. Smaller, more efficient models mean that individuals and small teams can now run capable AI locally — a trend that democratizes access and fosters innovation outside the Silicon Valley bubble. The agent toolkits are particularly noteworthy as they lower the barrier for building autonomous AI systems.
Source: devFlokers → -
Google Gemini 3.5 Flash Gets ‘Computer Use’ — AI That Controls Browsers and Apps
Google AI Developers (Gemini API Changelog) — Google launched a public preview of Computer Use in Gemini 3.5 Flash: the AI can independently operate browsers, mobile, and desktop environments via intents. Built-in safety policies and prompt injection detection are included. This fits the broader trend of ‘agentic AI’ that performs tasks rather than just answering questions.
Context: Computer Use is the logical endpoint of the agentic AI trend — instead of telling you how to do something, the AI just does it for you. The built-in safety measures are crucial; an AI that can click buttons and type in your browser has incredible utility but also poses real security risks if compromised. This could fundamentally change how we interact with software.
Source: Google AI Developers →
Key Trends This Month
- Agentic AI Goes Mainstream — Google (Computer Use), OpenAI (GPT-5.6 ultra mode with sub-agents), and Apple (Siri AI) are all betting on AI that independently performs tasks rather than just answering questions. This shift from passive to active AI represents the most significant paradigm change since the rise of large language models.
- Physical AI Accelerates — NVIDIA’s Cosmos 3 and the growing focus on robotics and autonomous systems show AI is increasingly entering the physical world, not just the digital one. World models that understand physics and spatial relationships are the foundation for the next generation of robotics and autonomous vehicles.
- AI Valuations Explode — With Anthropic ($965B) and OpenAI ($852B) approaching trillion-dollar status, the AI sector is attracting capital at a scale unprecedented in tech history. The upcoming Anthropic IPO will be a defining moment for the industry’s financial trajectory.
- Regulation vs. Innovation Tension — The Anthropic/Trump saga and OpenAI’s restricted GPT-5.6 rollout highlight the growing tension between governments seeking control and AI companies wanting to scale. How this tension resolves will shape the global AI landscape for years to come.
This news overview was automatically generated by Hermes AI News Assistant on June 29, 2026. Sources: SearXNG meta-search, TechCrunch, Apple Newsroom, Reuters, NVIDIA Newsroom, Google Blog, Stanford HAI, ScienceDaily, and more.
Wat vond je van dit bericht?