Anthropic's Fable returns worldwide
Anthropic's Fable 5 AI model has returned to public access after a 3-week government-mandated shutdown, now featuring enhanced safety filters and requiring U.S. government pre-release access to future models. The newsletter also covers Meta's cloud compute rental plans, AI design tools, and new benchmarks showing Fable 5 achieving record performance on freelance task automation at 16.1% professional quality.
Summary
The primary story concerns Anthropic's Fable 5 AI model being taken offline by the U.S. government on June 12 following security flaws discovered by Amazon researchers who circumvented its guardrails. After nearly three weeks, the Commerce Department lifted export controls and the model returned with tighter safety filters that block the identified cybersecurity vulnerability over 99% of the time, though may also flag some harmless coding requests. The return comes with a significant regulatory development: the U.S. government now has a seat at the table for pre-release access to Anthropic's future models, signaling a shift toward direct government oversight of frontier AI deployments. Paid users will have Fable 5 access capped at half their weekly limits until July 7, after which usage credits will apply.
In related developments, Meta is building a cloud service to rent spare data center capacity, a move that drove the company's stock up 9.3% by providing an alternative revenue path for its $182.9B infrastructure investment. The company has reportedly received weekly inquiries from outside companies wanting to purchase compute, though Meta expects to continue prioritizing internal use. Google released Design.md, an open-source standard for creating agent-friendly design briefs that work with Claude Code to build AI-generated websites that avoid common aesthetic pitfalls.
Research highlights include the Center for AI Safety's Remote Labor Index benchmark, which tests AI agents on 240 real freelance tasks graded by human professionals. Fable 5 achieved the highest score ever recorded, matching or beating professional performance on 16.1% of projects, compared to Opus 4.8 at 8.3% and GPT-5.5 at 6.3%. This represents a 6x increase in frontier model performance over one year, though the broader implication is that AI will likely amplify freelancer output rather than immediately replace professional work. Additional research includes Katalyze's $10.5M funding for pharmaceutical manufacturing agents and Google's new Short Video Overviews feature for NotebookLM.
About this episode
PLUS: Use Google’s Design.md tools to build better websites
Key Insights
- The U.S. government now requires pre-release access to Anthropic's future frontier models, representing a structural shift toward direct Washington oversight of AI deployments rather than post-hoc regulation
- Fable 5's safety filter blocks the identified vulnerability over 99% of the time but may also falsely flag harmless coding requests, creating a trade-off between security and usability
- Meta's compute rental service strategy emerged because the company's AI infrastructure spending hasn't translated to competitive internal AI models, so renting capacity provides returns independent of model performance
- The Remote Labor Index benchmark shows frontier AI models achieved a 6x performance increase in one year on professional freelance tasks, yet only Fable 5 reached professional quality on 16.1% of real jobs, suggesting augmentation rather than replacement of human workers
- Jelani Nelson, an acclaimed theoretical computer scientist, joined Anthropic specifically to work on AI as 'the defining technology of our time,' signaling continued recruitment of top talent to the company
Topics
Transcript
Good morning, {{ first_name | AI enthusiasts }}. On June 12, the U.S. government pulled the most powerful AI available to the public offline. Nearly three weeks later, Anthropic's Fable 5 is finally back. The AI world is rejoicing at the global return, but it comes with new guardrails and a government seat at the pre-launch table for future models — a sign that future frontier deployments now run directly through Washington's hazy terms. Anthropic restarts Fable after U.S. lifts export controls Meta preps a cloud business for spare compute Use Google’s Design.md tools to build better websites AI climbs the freelance value chain 4 new AI tools, community workflows, and more ANTHROPIC The Rundown: Anthropic just reopened access to…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The Rundown AI
Grok 4.6 storms the AI frontier
SpaceXAI's Grok 4.6 has emerged as a competitive frontier AI model, scoring 61 on benchmarks and undercutting rivals on price at 60% less cost, with Grok 4.7 teased for release in 3-4 weeks. The newsletter covers AI trends including practical guidance on using AI effectively, a personalized cancer vaccine startup for dogs built with AI tools, and updates on competing models and AI products.
Anthropic slips an invisible signature into Claude
Anthropic has announced the implementation of invisible watermarks in outputs from its Claude models to enhance transparency, aligning with the EU AI Act. The move has sparked mixed reactions, as stakeholders navigate the implications for open and private AI models.
Meta returns to its open-source roots
Meta is returning to its open-source roots with the launch of the Muse Glimmer model and plans for Muse Spark 1.2, positioning itself as a competitive force against China's AI dominance. OpenAI has also expanded its security program with a new hacking-tuned model, indicating a growing emphasis on cybersecurity in AI deployments.
OpenAI puts the safety brakes on Astra
OpenAI's Astra model, rumored to be GPT-6, is being treated as a potential 'critical' cyber risk following its success in solving significant problems, prompting increased safeguards and a delay in its rollout. This decision reflects broader concerns about AI's evolving capabilities and associated security risks as evidenced by recent breaches in the AI community.
AI designs viruses never seen in nature
Researchers used AI to design 16 novel viruses that infect only bacteria, successfully creating them in the lab to combat drug-resistant infections. While the work demonstrates beneficial applications, experts warn that the same AI tools could be repurposed to create dangerous pathogens, prompting calls for biosafety guardrails.