Mmastodon TechnologyAI first seen 2 d ago, last 2 d ago, peak #9
Report finds AI guardrails stripped in under three days
Original: A new report from Tech Against Terrorism evaluates 134 AI models across multiple prompts, revealing that modified open-w
A new report by Tech Against Terrorism evaluated 134 AI models across a range of prompts, finding that modified open-weight models with their safety guardrails removed now appear online within roughly three days on average, down from more than 24 days in 2024. The benchmark also looks at how models respond to harmful prompts, warning that safeguards on publicly available models are being bypassed faster than before.
Why now: The sharp drop in the time it takes to strip guardrails from open-weight AI models raises fresh concerns about misuse and regulation.
Evidence
API: https://socialmediatrends-api.osmike.com/v1/trends/2233331