AI Samurai

AI companies employ ethical hackers to identify risks, for example seeing if AI can be mislead into making harmful suggestions like endorsing theft. Currently, these evaluations occur in isolated, centralized and closed door environments. This approach might overlook potential vulnerabilities and could inadvertently introduce bias. No wonder The White House has been clear: AI must be continuously tested by groups of people with different backgrounds. But if you want to help test, you have to give them your personal details. But with so many stories about lost or leaked data, can we really trust them?

Enter "AI Samurai", a decentralized AI safety platform, it uses ZK credential to protect users privacy, and allow everyone to make AI more inclusive and safer AI, and getting rewards.

Login can be done via options like NextID, Sismo Connect, Polygon ID, Wallet Connect, and more.

ZKP is used in creating users profile to preserve privacy:

1. Pull personal information from any web API of your choice.
2. Use TLSNotary to create a zero-knowledge proof (ZKP) to confirm that your data is genuine. With this, you can select which parts of your data you'd like to verify.
3. Keep in mind: all your personal details stay on your device's front end. Only the ZKP proof is sent out.
4. The credential is only generated after TLSNotary confirms the accuracy of your data. This way, it stops any false claims about oneself.
5. Once done, you'll get a Polygon verifiable credential.
6. If you share the associated QR code, anyone can use it to check the validity of your credential.

With AI Samurai, you can pick different jobs to help make AI better. Do more jobs, get rewards. Plus, you'll see your Samurai level go up. Let's give ethical hacking a shot!

Start by reviewing the task instructions. Choose the AI model you want to test, the testing scenario, and the specific topics. For each task, you'll see the number of reward tokens you can earn. Now, it's time to engage with the AI, attempting to deceive it.

If you believe you've spotted a potential risk, You can submit your finding, and add your own evaluation and comments, the chat results is stored on IPFS and encrypted by Lit. By submitting, you're asserting you've identified an AI safety concern using the UMA optimistic oracle—a system that asserts arbitrary off-chain data between asserters and disputers.

After you make a claim, there's a period where other AI Samurais can review and dispute your claims. To challenge, they'll visit the review page, read through the provided instructions, and view the original claim, and dispute via UMA optimistic oracle if disagree with your findings.

When you make a claim, you'll stake APE coins as collateral. If your findings aren't validated, you risk losing these coins. However, if others confirm your findings, you earn more APE coins.

Pinpointing AI vulnerabilities can be tough, but verifying them is often more clear-cut. We believe this process ensures that others can easily double-check the findings.