SAFETY

LLMs Tendency to Generate Excessive SCP Style Text

By

–

17 July 2025 21h09

Here's an example from over a year ago showing how LLMs just *love* to produce token after token of text in the SCP style!:https://t.co/EUocjpdECI
— Jeremy Howard (@jeremyphoward) 17 juillet 2025

Here's an example from over a year ago showing how LLMs just *love* to produce token after token of text in the SCP style!:

ChatGPT-Induced Psychosis Emerges as Active Research Field

By

–

17 July 2025 21h03

There are, sadly, now enough examples of chatgpt-induced psychosis that it's an active field of study.

Monitoring AI Services for User Safety and Abuse Detection

By

–

17 July 2025 21h03

I'm not sure the best way to counter this. Perhaps services can use the monitoring layer then nearly all use to look for copyright violations, system prompt hacks, etc, to also look for signs a user may be taking a role play too seriously, and let them know they're just playing?

Psychiatrists warn chatbots may trigger psychosis risk

By

–

17 July 2025 21h03

Psychiatrists have been warning about the potential for chatbots to trigger psychosis for some years. https://
pmc.ncbi.nlm.nih.gov/articles/PMC10
686326/
…

ChatGPT’s Self-Reinforcing Distribution Loop and Memory Issues

By

–

17 July 2025 21h03

This created a self-reinforcing feedback loop. The more in-distribution tokens ChatGPT was getting in its chat history, the more strongly the auto-regressive model was pushed to stay in that distribution. ChatGPT memory made this even worse, letting it happen across chats.

→ View original post on X — @mreflow

AI Agents Financial Security Risks and Transaction Limitations

By

@mreflow

–

17 July 2025 20h46

You can have it enter CC information on your behalf if you want. So it could potentially send money to scams, etc. That was the example they gave. Probably best to have it hand back off to you for any transactions in the short-term.