A quieter news day with one outsized story: OpenAI shipped GPT-5.6 (Sol, Terra, Luna) on Jun 26 — but only to ~20 government-approved partners, after the White House asked it to limit the rollout over cybersecurity concerns.
It's the first time Washington has pre-cleared an American model launch, echoing the Fable 5 / Mythos 5 standoff and setting a precedent OpenAI openly dislikes. Elsewhere: a hard Google Veo 3 shutdown deadline lands Jun 30, and Cloudflare + beehiiv give publishers one-click AI-crawler controls.
TLDR
- OpenAI launches GPT-5.6 — Sol, Terra, Luna — but limited preview only, gated to ~20 US-government-approved partners (Jun 26) Trending Pricing Policy
- First-of-its-kind: White House (ONCD + OSTP) asked OpenAI to restrict GPT-5.6's release; OpenAI says this "shouldn't become the long-term default" (Jun 25–26) Policy Trending
- Google Veo 3 video models shut down Jun 30 — migrate to Veo 3.1 before the deadline (deadline Jun 30) Pricing
- Cloudflare + beehiiv give newsletter publishers one-click AI-crawler controls and a crawler-analytics dashboard (Jun 23) Customer AI Data
Action Items
Things your engineering team should look at or do this week.
- Migrate any Veo 3 integrations before Jun 30. Google is shutting down
veo-3.0-generate-001andveo-3.0-fast-generate-001on Jun 30; move to the Veo 3.1 model IDs (preview or GA) on Gemini API / Vertex AI to avoid a hard service break. (Google deprecations) - Don't plan against GPT-5.6 GA yet. It's a limited preview restricted to ~20 government-approved partners via API/Codex only — not in ChatGPT, no open API access. OpenAI says broad availability is "coming weeks," so keep roadmaps on GPT-5.5 / current models and budget for the new pricing tiers when it opens. (OpenAI)
- Watch the precedent, not just the product. A US pre-release review regime for frontier models (ONCD/OSTP) could gate future launches you depend on. Track it the way you'd track an upstream provider's release policy. (TechCrunch)
Models & Tools
- OpenAI previews GPT-5.6: Sol, Terra, and Luna (Jun 26) Trending Pricing — OpenAI began a limited preview of three new models: Sol (flagship, most capable), Terra (balanced everyday model, ~half the cost of GPT-5.5 with competitive performance), and Luna (fastest, most cost-efficient for high-volume routine tasks). The family is pitched as a frontier step on software engineering, computer use, knowledge work, scientific research, and cybersecurity. API pricing (per 1M tokensA small unit of data, such as part of a word, that AI models read and generate. Many AI services price by tokens in and out.NVIDIA ↗): Sol $5 in / $30 out, Terra $2.50 / $15, Luna $1 / $6. During the preview it's available only through the OpenAI API and Codex to a small set of trusted/approved partners — not in ChatGPT — with broader GA "in the coming weeks." (See Policy & Safety for why access is gated.) (OpenAI | VentureBeat | 9to5Mac)
APIs & Pricing
- Google Veo 3 video models reach end-of-life Jun 30 (deadline Jun 30) Pricing —
veo-3.0-generate-001andveo-3.0-fast-generate-001are deprecated and shut down on Jun 30, 2026 on both the Gemini API and Vertex AI. Migrate to the Veo 3.1 model IDs (veo-3.1-generate-preview/veo-3.1-fast-generate-preview, or the 3.1 GA models) before the cutoff to avoid disruption. (Google deprecations | Gemini pricing)
Industry & Business
- Cloudflare and beehiiv launch AI-crawler controls for publishers (Jun 23) Customer AI Data — The two companies integrated Cloudflare's Crawl Control into beehiiv, giving newsletter creators a dashboard showing which AI crawlers tried to access their content, which were blocked, and how much referral traffic each sent back — plus one-click allow/block per AI model and auto-updates as new crawlers appear (no manual robots.txt). All beehiiv users get beta visibility; Max customers can actively block crawlers. A concrete data point in the ongoing "who gets to train on my content" fight. (Search Engine Land | BusinessWire)
Policy & Safety
- US government asks OpenAI to limit GPT-5.6's release — a first (Jun 25–26) Policy Trending — The White House's Office of the National Cyber Director (ONCD) and Office of Science and Technology Policy (OSTP) asked OpenAI to restrict GPT-5.6 to ~20 government-approved partners while the administration builds a framework to test and evaluate new models' security, citing the model's cybersecurity capabilities. Reporting frames this as the first time the US has preemptively asked an American AI company to limit a launch before release. OpenAI is cooperating but pushed back publicly: "We don't believe this kind of government access process should become the long-term default." This rhymes directly with the Fable 5 / Mythos 5 export-control saga and signals an emerging pre-clearance regime for frontier models. (TechCrunch | Axios | CNN)