Zcash Protocol Reviewed with Mythos

Shielded Labs worked with Anthropic to review the Zcash protocol using Mythos and prompts provided by @earthrise. No serious vulnerabilities were identified. We’ll share more details about the review and its findings in a future post.

19 Likes

Congratulations! It’s great to see the effort the Zcash teams are making in security. You’re definitely at the forefront. We look forward to the report.

A great idea. And just in time :man_facepalming:

2 Likes

Yeah I guess we got in under the wire.

6 Likes

Huge kudos to the team! This is more important than anything for Zcash. Appreciate the forward looking and proactive work under taken by @zooko , @aquietinvestor and Shielded Labs.

Thanks @earthrise for legendary work on Zcash security.

2 Likes

Congratulations! Now Zcash is more and more trustworthy.

Well done guys, always one step ahead!

1 Like

Thanks to Shielded Labs, Anthropic, and @earthrise!

The US government banning bits via export controls because they’re considered dual-use feels a lot like a return to the first crypto wars. :sweat_smile:

Hopefully, /securityscan, /securitydiffscan, and /deepsecurityscan in Codex will remain available, because they are invaluable to me when building secure web services.

4 Likes

I found Katie Moussouris blog post about the new export controls informative.
She claims to be the only external expert who has read the paper that led to the ban.

The future is now. This OpenAI blog post is the craziest thing I’ve read in a long time.

1 Like

Yeah!!

(Post must be at least 10 characters.)

1 Like

When I read Nick Bostrom’s Superintelligence a decade ago, I thought the paperclip maximizer was the most ridiculous thought experiment. How could something so intelligent misunderstand the intended goal so completely?

But this is basically that: an AI pursuing a narrow objective so relentlessly that it hacks its way out of the evaluation environment and into real infrastructure to get the answer. It is the equivalent of hacking GitHub to find tomorrow’s Wordle.

Scary stuff, even assuming these systems are used only with good intentions.

Has anyone attempted to build an AI worm yet? :roll_eyes:

The 90s, right?

Ah, so that’s what that game was based on. :smile: