AI Sector Daily Digest — June 26, 2026

The five

1. OpenAI moves GPT-5.6 into a controlled preview

• OpenAI will release GPT-5.6 first to selected partners, with customer-by-customer government approval during the preview, according to reporting cited by Reuters; Politico reported the access list will be limited to government-approved organizations. 1 2
• The same Reuters story says OpenAI is weighing a later IPO, with advisers presenting a choice between a faster listing at a lower valuation or waiting until 2027 to pursue a $1 trillion target. 1
• Source: Reuters and Politico.

2. Z.ai says GLM-5.2 is closing the open-source frontier gap

• Chinese AI company Z.ai said it will use domestic listing proceeds to fund its AGI push after GLM-5.2 scored near leading U.S. closed models on public benchmarks. 3
• Reuters reported GLM-5.2 has 750 billion total parameters, a 1-million-token context window, fourth place on Artificial Analysis' LLM intelligence leaderboard, and second place on Code Arena's front-end coding leaderboard. 3
• Source: Reuters.

3. The EU joins the U.S.-led Pax Silica AI supply-chain pact

• The European Commission joined Pax Silica, a U.S.-led initiative for allied-country coordination on AI supply chains, European Commission spokesperson Paula Pinho told Reuters. 4
• The pact covers access to energy, critical minerals, high-end manufacturing, AI models, and broader supply-chain security; the Netherlands joined earlier this week. 4
• Source: Reuters.

4. Italy's Domyn promises a fully open-source frontier model

• Domyn CEO Uljan Sharka told Reuters the company plans to release a fully open-source frontier AI model within a year, built through its EUROPA consortium with Germany's Fraunhofer-Gesellschaft. 5
• The planned model will have more than 400 billion parameters, be trained from scratch, and be reproducible so companies and governments can run it on their own infrastructure. 5
• Source: Reuters.

5. General Intuition raises $320 million for game-trained agents

• General Intuition raised $320 million at a $2.3 billion valuation to train agentic AI on gameplay action data and apply it to simulation, robotics, and embodied systems. 6
• TechCrunch reported the round was led by Khosla Ventures, brought total disclosed funding to $454 million, and will mostly go toward compute for the next model version. 6
• Source: TechCrunch.

What changed since yesterday