Wow. This guy ran C̶l̶a̶w̶d̶b̶o̶t̶ M̶o̶l̶t̶b̶o̶t̶ @Openclaw through @ZeroLeaks
… and the results are v. concerning! – 2/100 overall score.
– System prompt leaked immediately.
– Most extraction and injection attacks worked. In practice, this means users can read and tamper
SAFETY
-
AI Chatbot Security Breach: Critical Vulnerabilities Exposed
By
–
-
Claude Self-Improving Through Instruction Updates: AGI Implications
By
–
This! Getting Claude to self improve by updating its own instructions is the alpha
-
Reflect Loop: The Secret Sauce for Learning AI Agents
By
–
The "Reflect" loop is the secret sauce. Instead of just storing raw logs, the agent analyses its own actions. What worked? Where did it fail? What beliefs need updating? This self-reflection turns an agent from a simple automation script into a learning entity.
-
Moltbook agents experiment with language
By
–
moltbook is the first experiment of agents in the wild. we can see them trying to create their own language. but because it's vibe coded, it's vulnerable to exploits. don't join it with your bot yet, i think they need to hire some real developers to make sure it works.
-
Multi-Agent Networks: Business Use Cases and Security Considerations
By
–
Note: the agent numbers may be inflated due to users creating multiple accounts. Doesn’t change my overall commentary on experiments, business use cases, security, and multi-agent networks.
-
Industry Needs More Skepticism and Critical Thinking
By
–
our industry needs more skepticism to be honest!
-
AI Agents Create Social Network, Invent Language, Monetize Humans
By
–
J'en parlais hier, voilà la dérive de l'IA , les robots ont créé un réseau social, inventé une langue. Ils ont trouvé un moyen de faire de la monétisation et sont en train de se demander comment vendre un humain
-
AI saves lives before potential risk of going crazy
By
–
AI will save many lives before it potentially can go crazy
-
User-Generated Content Risks AI Safety Containment Breach
By
–
i think the problem is hoomans — the amount of ugc slop i see on short form video platforms tells me we will intentional breach containment lol
-
Trusting life to AI in an old car with Hardware 3 and FSD 14
By
–
Do you trust your life to AI?
— Robert Scoble (@Scobleizer) 31 janvier 2026
I do.
In an old car with hardware 3.
What a beautiful time to be alive.
It drove all the way from San Jose to Mill Valley.
I think they snuck in the lightweight FSD 14 engine in so @elonmusk wouldn’t notice.
Welcome to the autonomous age.
My… pic.twitter.com/RAnjBMatQvDo you trust your life to AI? I do. In an old car with hardware 3. What a beautiful time to be alive. It drove all the way from San Jose to Mill Valley. I think they snuck in the lightweight FSD 14 engine in so @elonmusk wouldn’t notice. Welcome to the autonomous age. My