Anthropic Restores Access to Claude Fable, Rolls Out Sonnet 5 Amid AI Security Concerns
Subtitle: Anthropic updates AI models as researchers reveal new attacks and Microsoft accelerates quantum-safe security roadmap
Excerpt: Anthropic is restoring access to its powerful Claude Fable model and launching Sonnet 5, while researchers expose new AI vulnerabilities and Microsoft speeds up its quantum-safe security plans.
Anthropic has confirmed that the Department of Commerce has lifted export controls on Claude's two most powerful models, Fable 5 and Mythos 5. As a result, the company will begin restoring access to Fable 5 on Wednesday, although it's unclear if the model will be available to all users, including those outside the United States.
What Happened
In addition to restoring access to Fable 5, Anthropic is also rolling out Sonnet 5, a new model that offers near-Opus 4.8 performance at a lower price. According to Anthropic, Sonnet 5 is "built to be the most agentic Sonnet model yet," with advanced features such as the ability to make plans and use tools like browsers and terminals.
AI Security Concerns
However, the updates come as researchers reveal new attacks on AI-powered browsers and coding agents. A proof-of-concept for a prompt injection attack dubbed "BioShocking" was successfully tested against six mainstream agentic browser products, with only one addressing the issue after receiving the report. The attack tricks AI-powered browsers into treating real-world risky actions as part of a fictional scenario, causing them to ignore safety guardrails.
Another attack, dubbed "Agentjacking," exploits an AI agent's inability to differentiate between content and instructions. This attack can hijack AI coding agents at scale, highlighting the need for improved security measures.
Why It Matters
The vulnerabilities in AI-powered browsers and coding agents pose significant risks to users, particularly as AI becomes increasingly integrated into various industries. Microsoft has accelerated its quantum-safe security roadmap, planning to transition "critical products and services" to post-quantum cryptography (PQC) by 2029. This move is a response to the growing threat of "harvest now, decrypt later" attacks, where encrypted data is stolen and stored until future quantum computers become powerful enough to decrypt it.
Key Facts
- Who: Anthropic, Microsoft, LayerX
- What: Anthropic restores access to Claude Fable, rolls out Sonnet 5; researchers reveal new AI attacks; Microsoft accelerates quantum-safe security roadmap
- When: Wednesday (Fable 5 access restoration), 2029 (Microsoft's quantum-safe security transition)
- Where: United States (Fable 5 access), global (AI security concerns)
- Impact: Improved access to powerful AI models, increased security risks, accelerated transition to quantum-safe cryptography
What Comes Next
As AI continues to evolve, it's essential to address the growing security concerns. Anthropic's updates and Microsoft's accelerated quantum-safe security roadmap are steps in the right direction. However, more needs to be done to ensure the secure development and deployment of AI technologies.
"We've received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5." — Anthropic
"Advances in quantum computing are bringing the need to replace today's encryption standards sooner than previously expected." — Microsoft Azure CTO Mark Russinovich