Happy Thursday! Last week Anthropic launched cheaper Claude Sonnet 5 and saw its Claude Fable 5 model return globally after an export ban was lifted, while OpenAI unveiled GPT-5.6 to compete directly but faced government-imposed release restrictions.

NEW GENERATIVE AI LAUNCH

Claude Fable 5 Returns Globally After Export Ban Lifted

Anthropic is restoring worldwide access to Claude Fable 5 after the U.S. Department of Commerce withdrew emergency export controls issued on June 12, 2026. The restrictions had forced Anthropic to suspend global access to both Fable 5 and its cybersecurity-focused counterpart, Claude Mythos 5, shortly after their launch. Fable 5 is now available across Claude Platform, Claude.ai, Claude Code, and Claude Cowork. Anthropic is also working to restore access on Amazon Web Services, Google Cloud, and Microsoft Foundry as quickly as possible, though full cloud hyperscaler availability remains pending.

NEW GENERATIVE AI LAUNCH

Anthropic Launches Cheaper Claude Sonnet 5 Before IPO

Anthropic has released Claude Sonnet 5, promising near-flagship AI performance at mid-tier prices. The model becomes the default for Free and Pro plan users, with introductory API pricing of $2 per million input tokens and $10 per million output tokens through August 31, rising modestly after that. The launch comes as Anthropic races toward a high-profile IPO that will test whether sky-high private AI valuations hold up under public market scrutiny. Sonnet 5 is priced well below the company's premium Opus 4 model, signaling a clear push to attract cost-conscious enterprise developers.

NEW GENERATIVE AI LAUNCH

Trump Eases Ban on Anthropic's Mythos 5 Model

The Trump administration is partially lifting its ban on Anthropic's cybersecurity AI models, allowing Mythos 5 to be accessed by more than 100 U.S. government agencies and companies. The directive, signed by Commerce Secretary Howard Lutnick, also permits non-American employees at those organizations to use the model. The ban was imposed after security researchers allegedly bypassed the models' guardrails. Fable 5 remains restricted under the new directive. Anthropic confirmed it is restoring access for organizations that operate and defend critical infrastructure, while continuing to work with the government to expand access further.

NEW GENERATIVE AI LAUNCH

OpenAI GPT-5.6 Targets Anthropic's Claude Mythos 5

OpenAI has launched GPT-5.6, a three-tier model series comprising Sol, Terra, and Luna, designed to compete with Anthropic's Claude Mythos 5. The lineup introduces two new modes: a max setting for deeper reasoning and an ultra mode enabling parallel subagents. Sol scored 91.9% on TerminalBench-2.1 with ultra mode enabled, edging past Claude Mythos 5's 88%. Sol is priced at $5 per million input tokens and $30 per million output tokens, with Terra and Luna offering significant discounts. OpenAI is initially limiting access to trusted partners at the U.S. government's request, with general availability expected within weeks.

AI REGULATORY UPDATE

OpenAI Limits GPT-5.6 Release Under Government Pressure

OpenAI is restricting its new GPT-5.6 lineup — flagship Sol, balanced Terra, and budget Luna — to a small group of government-approved partners after a Trump administration request. The move follows the administration's effective ban on Anthropic's Fable 5 model, raising concerns about government overreach in AI releases. OpenAI pushed back, calling the arrangement a short-term step and warning that restricting advanced AI tools harms developers, enterprises, and global partners. The company plans broader availability soon and is working with the administration on a repeatable framework for future model releases.

GENERATIVE AI COMPETITION

OpenAI and Anthropic Face Same Government Bottleneck

The U.S. government is tightening control over frontier AI model releases, with OpenAI's GPT 5.6 facing a customer-by-customer approval process similar to the one that has kept Anthropic's Mythos in limited preview for months. The delays threaten the economic viability of costly new systems and could slow the broader data center buildout. The crisis has exposed a deeper problem: the government lacks clear criteria or capacity for evaluating AI models, yet real concerns around cybersecurity and biorisk exist. Experts argue the industry must stop treating regulation as a competitive weapon and instead pursue collective action to establish a workable release framework.

AI REGULATORY UPDATE

OpenAI Offers Trump Admin 5% Ownership Stake

OpenAI has proposed giving the US government a 5 percent ownership stake as a strategy to ease tensions with the Trump administration and counter growing public opposition to AI, according to the Financial Times. CEO Sam Altman argued that giving the public a direct financial interest in the company would be the most effective way to share the benefits of the AI boom.

NEW GENERATIVE AI LAUNCH

Google Launches Faster, Cheaper AI Image Generator

Google has released Nano Banana 2 Lite, a faster and more affordable AI image generator capable of producing images in four seconds at $0.034 per 1,000 images. The model is designed for high-volume workflows and replaces the original Nano Banana, now labeled a legacy model. Google also announced a wider release of Gemini Omni Flash, priced at $0.10 per second of video output, and unveiled Omni Product Studio, a demo app that transforms static images into cinematic e-commerce videos. Both tools are available via Google AI Studio and the Gemini API.

NEW GENERATIVE AI HARDWARE

OpenAI Teases Codex Hardware Device for July

OpenAI is set to unveil a new hardware device tied to its AI coding tool, Codex, on July 15th. A teaser video posted to X shows a square-shaped device with several buttons, hinting at upgraded shortcuts for the platform. This device appears separate from the secretive AI hardware OpenAI is reportedly developing with other partners. Few details have been revealed beyond the brief teaser and its July 15th release date.

NEW AI STARTUP FUNDING

Amazon Launches $1 Billion AI Deployment Engineer Org

Amazon Web Services has launched a new internal organization of forward-deployed engineers focused on AI integration, committing $1 billion in internal resources. The team will embed directly within client companies to deploy AI agents, with a focus on speed and building lasting in-house capabilities for customers. The forward-deployed engineer model, pioneered by Palantir, has grown popular for managing complex AI rollouts. Amazon joins OpenAI and Anthropic, which recently launched similar ventures valued at $4 billion and $1.5 billion respectively, though those were backed by private equity partners rather than internal funding.

Upcoming Data & AI Conferences

Ai4 2026

Tue, Aug 4 - Thu, Aug 6, 2026 | Las Vegas, NV

Curated with AI by EMC2 AI

Brought to you by Estevan McCalley

Share this issue

Follow Estevan

Like this? Subscribe for the weekly AI roundup.