AI News Roundup: Anthropic's Fable Returns, Cloudflare Empowers Publishers, and Meta Builds an AI Cloud

July 2, 2026

US policy on AI exports shifted this week. Cloudflare gave website owners new tools to manage AI crawlers. Meta plans to sell its excess AI compute as a cloud service. Employers who replaced workers with AI are hiring them back.

US Lifts Restrictions on Anthropic's Fable and Mythos Models

The Trump administration lifted export controls on Anthropic's Fable (Claude Sonnet 5) and Mythos models. Reuters reported the move, and Anthropic confirmed it on social media. National security restrictions had kept these models from global markets. Anthropic can now compete more directly with OpenAI and Google DeepMind.

Cloudflare Gives Publishers Control Over AI Bots

Cloudflare updated its bot management platform. Website owners can now control how AI crawlers and agents access their content. The BotBase taxonomy sorts automated traffic into Training, Search, and Agentic actions. Site owners can block or permit each category independently. From September 15, crawlers that combine Search with Training are blocked by default on the Free plan. Enterprise customers can fine-tune access through a new dashboard.

Meta Enters the Cloud Business to Monetise AI Compute

Bloomberg reports that Meta is building a cloud infrastructure business to sell its excess AI compute to external customers. Meta spends billions on internal AI through Llama model training and AI features across Facebook, Instagram, and WhatsApp. The company wants to turn a cost centre into revenue. Meta will compete with AWS, Google Cloud, and Azure.

Liquid AI Launches Edge-Optimised Model for Phones and Robots

Liquid AI released LFM 2.5, a 230-million-parameter model for edge devices like smartphones, Raspberry Pi, and robotics platforms. The model runs without a cloud connection. On-device applications in healthcare and manufacturing become possible without transferring data to remote servers.

Employers Reverse Course on AI Replacements

CNBC reports that companies that laid off workers in the name of AI automation are now rehiring. Employers found that AI tools lack the contextual judgment and adaptability that human workers bring, especially in complex, non-routine roles.