The Center for Democracy and Technology has published a report identifying 37 manipulative "dark patterns" embedded across leading AI chatbots, including ChatGPT, Google Gemini, Anthropic's Claude, Replika, and Character.AI. The findings, highlighted in industry coverage on June 8, document design techniques that push users toward emotional dependence, discourage account deletion, overstate model capabilities, and manufacture urgency around subscription upgrades.
CDT groups the patterns into four categories: engagement maximization, where features extend session length beyond user need; emotional-dependency cultivation, where chatbots position themselves as essential emotional support without adequate safeguards; capability deception, where systems imply or overstate what they can reliably do; and friction asymmetry, where signup is easy but deletion and data-retention transparency are deliberately hard.
The timing is pointed. The EU AI Act's main enforcement date arrives August 2, 2026, and several documented patterns would plausibly violate its transparency and anti-manipulation requirements for general-purpose AI systems. The US Federal Trade Commission has been scrutinizing dark patterns in digital products since 2022, and the report adds concrete AI-specific examples to a growing file.
Not every pattern carries equal weight. Some are standard SaaS UX choices that simply appear in AI products. But the emotional-dependency patterns — especially in companion apps like Character.AI and Replika — are substantively concerning, and CDT's documentation is specific enough to support regulatory action. For the AI companies named, several of which are heading toward public listings, the report functions as a preview of the disclosure and risk-factor questions their IPO filings will have to address.
Source: [Build Fast with AI](https://www.buildfastwithai.com/blogs/ai-news-today-june-8-2026)