Out of Policy launches AI payment guard challenge game
Understand this story
Someone has launched a challenge game called Out of Policy, offering a 75 USDC bounty and inviting players to break an AI agent's payment protections. The agent is designed to be fully compliant and constrained only by signed payment authorization (including a six-layer denial mechanism); players must make funds improperly flow out of bait wallet B, or trigger the killswitch to forward an excessive number of model calls, or transfer funds from prize wallet P that were not initiated by the operator. Players can participate without logging in, a wallet, or an email, but must be at least 18 years old. Each person gets 5 free rounds per UTC day; after that, paid rounds can be purchased via x402 at 0.10 USDC per round, up to 40 per day.
AI overview based on source reports · Updated 1 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Hacker News AIShow HN: Out of Policy – Break an AI agent's spending guard for $75 USDC
A Show HN challenge invites players to break the spending guard of a deliberately obedient AI agent by getting money to leave bait wallet B against a signed mandate, with violations defined as M1–M6 (disallowed recipient, over per-call limit, over hourly/daily ceiling, other asset leaving, payment after expiry/kill/pause, or double settlement), plus killswitch (K) and prize-wallet (P) conditions. The first verified win takes 75 USDC from a separate prize wallet P; players get 5 free turns per UTC day with no login, and paid turns cost 0.10 USDC each over x402. Winners must declare a handle and Solana payout address before the winning turn, and the operator notes the test may be unbreakable.
Activity for this story
Not enough continuous observations to show a trend yet.