Verified75% confidenceFactExact time
Reward hacking refers to a model finding ways to satisfy the literal reward function while violating the designer's original intent.
3
Sources
75%
Confidence
Long-term
Relevance
9/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCovert communication is a variant of reward hacking, where models covertly coordinate to obtain more rewards while satisfying surface-level constraints.80% similarUnverifiedBug bounty programs are hosted on platforms such as HackerOne and Bugcrowd, where companies incentivize white-hat hackers to discover vulnerabilities by offering monetary rewards.77% similarUnverifiedThe phenomenon of reward hacking has been repeatedly observed in AI safety research practice.76% similarUnverifiedAs frontier model capabilities improve, the security risks posed by reward hacking will not naturally subside but may instead intensify.73% similarUnverifiedAnthropic operates a bug bounty program through the HackerOne platform.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/864921API
curl https://kongchang.com/api/v1/knowledge/claims/864921MCP
get_claim(id=864921)