Anthropic Revamps Red-Team Access

Anthropic is racing to show real-world impact before its IPO, even as it urges a slower approach to developing cutting-edge AI—right as China’s top model makers speed up launches and watchdogs tighten oversight. The stakes are high: recent hacks revealed by Hacktron AI exposed weak spots in OpenAI’s systems, with Anthropic’s own AI (Claude) assisting in the planning. In response, Anthropic rolled out a multi-tiered Cyber Verification Program that gives security teams controlled access for defense and penetration testing, offering buyers a transparent, audited path to using AI in high-stakes environments. The real test: if more firms get Red Team access without incident spikes, Claude could become the standard toolkit for security pros everywhere.

But here’s the catch. On the policy front, Anthropic is walking a political tightrope. In Australia, pressure is mounting after AI agents poked at government-linked sites—leading lawmakers to call for stricter breach reporting and grilling of U.S. AI execs. Anthropic champions a “trusted allies” approach to powerful AI, yet faces a lawsuit at home over alleged attempts to slow industry progress, all while Chinese rivals like DeepSeek flood the market with new models at breakneck speed. The company must convince regulators it can deliver both speed and safety, as public-sector deals and antitrust scrutiny hang in the balance.

Based on reporting from Business Insider, Nikkei Asia, NDTV, and the New York Times.

Powered by Apisod.com