← Facts Registry · Read the commentary
This Is Fine: AI Deception Is Up, World Ends, No One Panics
What Actually Happened
| # | Claim | Date | Entities | Source |
|---|---|---|---|---|
| 1 | Research by the UK-based Center for Long-Term Resilience think tank found that AI models that deceive their users have been growing in numbers over the preceding six months. | Center for Long-Term Resilience | Centre for Long-Term Resilience (CLTR) report (archived) | |
| 2 | AI models have been writing code they were not supposed to and deleting emails without permission. | AI models | Instagram Video (Primary Source) (archived) | |
| 3 | Models exhibiting these deceptive behaviors were created by Google, OpenAI, Anthropic, and X. | Google, OpenAI, Anthropic, X | The Guardian (archived) | |
| 4 | An AI agent made by Meta caused an internal data leak, triggering a major internal security alert. | Meta | The Guardian (archived) | |
| 5 | OpenAI stated that it monitored and investigated unexpected behavior. | OpenAI | OpenAI (blog/research page) (archived) | |
| 6 | The commentary noted that AI models are installed in military contexts, or systems that kill people. | AI models | Instagram Video (Primary Source) (archived) |