Go-Kart Safety Won't Protect a Race Car
Is today's AI safe because the safeguards work, or because the AI is still limited?
METR reviewed part of Anthropic's risk report and agreed the risk was low, but mainly because of the models' limited abilities. Disclosure: this show is made with help from Anthropic's Claude.
Full episode: https://youtu.be/6yJ2T6-ZhOg
#AI #AISafety #Anthropic
AI Coalition Network: https://ai-coalition.net
More Shorts
All videos
AI Doesn't Have to Hate You to Hurt You
Nov 5, 2026
0 views
An AI Used Fake Identities During a Safety Test
Nov 5, 2026
1 view
It's Not Your Daughter on the Phone
Nov 5, 2026
0 views
AI Agents Are Getting Shared Skills
Nov 4, 2026
3 views
Should Government Trust AI From Other Countries?
Nov 4, 2026
4 views
A Confident AI Answer Isn't Always a Correct One
Nov 4, 2026
3 views
AI That Runs on Your Own Computer
Nov 4, 2026
1 view
What If AI Decided Where Your Calls Go?
Nov 3, 2026
3 views