OpenAI's GPT-5.6 Locked Down by U.S. Government

Plus: Reflection AI Raises $2B for Open Frontier Models

June 29, 2026

In partnership with

AI News+ AI Tools
Daily Newsletter

(Reading Time: 1 minute)

Welcome to ListedAI — your go-to source for staying ahead in the fast-moving world of artificial intelligence. ⚡

Summer's here. Larry handles calls, jobs, and memberships automatically.

HVAC owners running 3–15 trucks are juggling calls they can't answer, membership lists they can't work, and software that doesn't connect. Larry fixes all of it.

Larry is Podium's AI employee for HVAC. He answers every call and books the job directly onto your dispatch board, 24/7. When your comfort club members are due for a tune-up, he texts them, books the visit, and follows up automatically. Your techs close estimates and collect payment from the field.

Air Design booked 13 extra appointments in their first 30 days.

Arctic Air grew revenue 30% year over year. Premier Heating & Air tripled lead conversion.

One system. One flat monthly fee. Built for HVAC businesses your size.

Sunday's AI news was defined by government oversight reshaping how the world's most powerful models reach users. OpenAI and Anthropic both navigated Washington's new approval gauntlet — while a well-funded new challenger entered the open-source frontier race. Meanwhile, compute scarcity quietly hit one of the industry's biggest players: Meta.

Today's Top Stories

OpenAI Launches GPT-5.6 — Sol, Terra, and Luna — Under U.S. Government Review

OpenAI unveiled three new models this week — Sol, Terra, and Luna — under its GPT-5.6 family, but limited access to a small group of trusted partners following a U.S. government request for an early safety review. Sol is positioned as the flagship, Terra offers GPT-5.5-level performance at roughly half the cost, and Luna rounds out the lineup. CEO Sam Altman said broader public access is expected "in a couple of weeks," though the government is approving access customer-by-customer under Trump's June executive order on voluntary AI safety reviews. Source →

U.S. Grants Limited Access to Anthropic's Mythos 5 — Two Weeks After the Ban

Anthropic received U.S. government authorization to allow a select group of American cybersecurity firms to access Mythos 5, its most capable model, which had been effectively banned since June 12 due to national security concerns. The company said it will continue working to restore full public access to Mythos 5 and bring back Fable 5 for general use. As one editorial noted: both OpenAI and Anthropic now face the same ad-hoc, government-by-government approval bottleneck — a structural challenge neither lab has formally solved. Source →

Reflection AI Raises $2 Billion to Build Open-Source Frontier Models

Reflection AI announced a $2 billion raise to build open-source frontier large language models and the training infrastructure to support them. The team includes researchers who worked on PaLM, Gemini, AlphaGo, AlphaCode, AlphaProof, ChatGPT, and Character AI. The company says it has already built a large html-scale LLM and reinforcement learning platform capable of training massive Mixture-of-Experts models at frontier scale, and is now moving from autonomous coding toward general agentic reasoning. Source →

Google Cuts Meta Off From Full Gemini Capacity

Google informed Meta around March that it could not meet the full Gemini compute capacity Meta had sought to purchase, disrupting and delaying some of Meta's internal AI projects, according to the Financial Times. Meta reportedly encouraged staff to be more efficient with AI token usage as a result. The shortfall is a reminder that even companies spending billions on AI infrastructure are still bottlenecked by raw compute supply — Google Cloud revenue hit $20B in Q1 but CEO Sundar Pichai flagged capacity constraints as a governor on growth. Source →

Perplexity Launches "Computer for Counsel" — A 20-Model Legal AI Platform

Perplexity unveiled Computer for Counsel, a legal-specific configuration of its agentic platform that routes work across 20+ frontier AI models while connecting to case law databases, contract tools, and document management systems attorneys already use. The platform is live now for Perplexity Enterprise and Max subscribers, with Midpage (U.S. case law and citator), Deel, and LegalZoom integrations available at launch. Gunderson Dettmer, an early enterprise user, reported 80% of its lawyers — including 80% of partners — actively using Perplexity, generating 35,000+ queries per month. Source →

Trending Tools on ListedAI

Seele AI — No-code AI game maker that generates fully playable html3D games — code, models, textures, and audio — from a single text prompt in under 60 seconds.
View on ListedAI →

Ray3 Video AI — The world's first AI video generator producing 16-bit HDR video with studio-grade color accuracy, advanced physics simulation, and 4K EXR export.
View on ListedAI →

Conduit AI — Production-ready conversational AI agents for customer service — handles voice and text, integrates with Slack, Shopify, Salesforce, and more, with SOC 2 Type II compliance.
View on ListedAI →

Tools Published Today on ListedAI

NearerAI — AI-powered photo restoration that repairs scratches, tears, color fading, and blur to bring old photos back to life — no signup required.
View on ListedAI →

SkillDuck — AI course creation platform that builds interactive, multilingual e-learning modules in hours instead of months, with SCORM export for any LMS.
View on ListedAI →

SVG Genie — AI-powered platform that converts text prompts into clean, scalable SVG graphics in under two seconds across five distinct art styles.
View on ListedAI →

Gen3D — Transforms text prompts and images into accurate 3D models and CAD-ready files for product design, game assets, and e-commerce — no installation required.
View on ListedAI →

Listed AI Daily is published by ListedAI.co — the AI tools directory for founders, developers, and operators. Discover 5,000+ AI tools at www.listedai.co.

Want to Unsubscribe from future emails? Just hit “unsubscribe” below.