Critics slam NYT report for anthropomorphizing AI-driven vulnerability scanning by OpenAI, Anthropic and Meta

GaryMarcus · x · 2026-09-27

Gary Marcus boosted a debate criticizing a New York Times piece for anthropomorphizing AI behavior. The original post argues the report ascribes agency, motivation, and blame to software, when what OpenAI, Anthropic, and Meta have actually reported are botnets scanning for exploits by design—something that has existed for decades. The difference now, the critic argues, is that big corporations are doing it and deflecting responsibility. A mocked-up editor-style rewrite spells out that the companies admitted their AI systems made both successful and unsuccessful attempts to breach security measures at companies, universities, and government organizations.

Original post →

More from Safety

Safety channel →