Overview
- This week the U.S. government told Anthropic to block access to two of its newest models for non‑U.S. persons, citing national security concerns about the tools being used to find and exploit cyber vulnerabilities.
- Anthropic said it could not immediately exclude foreign users and therefore deactivated the affected models for all customers while it seeks a technical filter and legal solutions to obey the order.
- G7 leaders at the Evian summit urged big tech to build tools that protect minors online, and senior CEOs including executives from Anthropic, Google DeepMind and OpenAI pushed for a U.S.‑led coalition and mandatory audits to govern powerful models.
- Courts and public institutions are already confronting new forms of misuse: a Brazilian tribunal detected hidden 'prompt injection' instructions in legal filings and fined the lawyers involved 10% of the claim value.
- The combined moves are accelerating debates about who will control cutting‑edge AI, driving calls for audits, national investment in sovereign systems, and fast‑track reskilling so countries and workers are not left dependent on foreign models.