1/ i’m really impressed by the careful work we did to make Muse secure. you can read our technical post here:

2/ we also have a public bug bounty program. If you can find a security hole or impactful prompt injection in Muse, and disclose it responsibly, you can get up to $300,000. read more on our bug bounty page:

It's very secure!

We made it private, too!

セキュリティ作業も大事だけど、子供のガムテープ対策にも真剣です😂

Waitlisted 😩😩 drop us some invite codes g i was an scale customer

Is this how you guys start earning public trust for Muse?

敢把安全细节和漏洞赏金摆上台面，这才是硬底气。

@dps Well articulated security architecture ftw! You really can’t trust a personal agent without this.

@dps trust you’ll get it right this time!!!

Incredible product and launch

I am not impressed. When I connected my email it looked at the traffic ticket a friend of mine forwarded to discuss his situation as my traffic ticket and alerted me. There is no way to connect to whatsapp which is also a facebook product. I will continue to explore.

A public bounty for prompt injection is the right call, but the payout is downstream of the real fix. You cannot test your way out of a tool that can send, pay, or delete. Keep those behind an approval and the model has less to get wrong.

Why the link redirect to a security blog and and not the download page?

does muse test indirect prompt injection, since sandboxing alone won't stop authorized data leaks

Impressing oneself

The Sentinel architecture is the interesting part to me. If an agent action matters enough to gate, it may also matter enough to prove — what was authorized, what actually executed, and what changed as a result.

whats the biggest threat model you weighted for the agent, prompt injection through the pages it visits?