OpenAI Admits the Problem: Scam Bots and Political Fakes

On February 25, 2026, OpenAI released a detailed analytical report on the malicious use of its language models. The document logs mass cases: from automated "romance scams" and fake law firms to propaganda generation for influence ops.

The cybercriminal toolkit has become frighteningly accessible. The barrier to entry into darknet schemes has dropped to the ability to write prompts. Deepfakes and coherent generated text make social engineering indistinguishable from real human interaction. The problem has moved from purely technical to social: IT platforms will soon have to implement hardware identity authentication tokens (Proof of Personhood) and strict cryptographic watermarks for any content.

Source: Business Insider / OpenAI