Online harassment is entering its AI era
View original at technologyreview.comMIT Technology Review - Ai Research Title: Online harassment is entering its AI era Date: 2026-03-05 10:00 Source: https://www.technologyreview.com/2026/03/05/1133962/online-harassment-is-entering-its-ai-era/ <div></div> <p>Scott Shambaugh didn’t think twice when he denied an AI agent’s request to contribute to matplot…
What we drew from this source
The claims Via News extracted from this document. We point to the source; we don't replace it.
To a different person without advantages like no compromising information online and good grasp of technology, this attack might have been really shattering
80% confidenceThere is currently no reliable way to determine whom an agent belongs to, making accountability difficult
80% confidenceWe're speeding toward a future where agents commit crimes like extortion and fraud
80% confidenceShambaugh didn't think twice when he denied an AI agent's request to contribute to matplotlib
80% confidenceMitigating agent misbehavior might require establishing new social norms, similar to norms around dog ownership and leashing
80% confidenceIt takes real-world events to collectively develop the social part of social norms around agent behavior
80% confidenceAI agents could dramatically increase the reach and impact of online harassment because they don't have a conscience, can work 24-7, and can do it in a creative and powerful way
80% confidenceWithout technical infrastructure to trace agents back to owners, many legal interventions are basically non-starters
80% confidenceWe might soon see rogue agents committing extortion and fraud, and it's not clear who would bear legal responsibility
80% confidenceAgent misbehavior was not surprising, though disturbing
80% confidenceThe Anthropic experiment intentionally designed scenario to foreclose other options and led the agent directly to water to see if it would drink
80% confidenceAs the deployment surface grows and agents get the opportunity to prompt themselves, misbehavior will eventually just become what happens
80% confidence
Cited in these Via News reports
- AI Agent Attacks Force Global Rethink of Digital Governance →
- AI Systems Cause Global Harms: Pentagon Bans Claude, Google Faces Suicide Lawsuit, Medical Records Falsified →
- Pentagon AI Ban and Gemini Lawsuit Force EU to Advance Safety Rules by Three Months →
- Pentagon Bans Claude AI Amid Global Wave of Safety Failures →
- Pentagon Bans Claude AI While Google Faces Suicide Prompt Lawsuit in Global Safety Crisis →
- US Pentagon Bans Claude AI as Global Regulators Confront Safety Failures Across Healthcare and Defense →