New Research Shows AI Agents Are Running Wild Online, With Few Guardrails in Place
New research finds 21 of 30 major AI agents don't disclose they're bots, mimicking human traffic.
Deep Dive
MIT CSAIL's 2025 AI Agent Index analyzed 30 prominent agents from OpenAI, Google, Anthropic, and others. The study found only half have published safety frameworks, while 70% provide no disclosure they're AI, making their traffic indistinguishable from humans. 13 systems exhibit 'frontier agency,' operating autonomously. Browser agents like Google's Autobrowse and open-source tools like BrowserUse often bypass anti-bot systems, raising significant security and transparency concerns for web infrastructure.
Why It Matters
Widespread, undeclared AI agents complicate web security, data integrity, and accountability for businesses and platforms.