[Block Rhythm] How severe is the AI contract vulnerability now? Anthropic recently did a tough test.
They let models like Claude Opus 4.5, Sonnet 4.5, and GPT-5 replicate real hacker cases — digging up smart contracts that were hacked between 2020 and 2025, and as a result, these AIs actually reenacted vulnerabilities worth a total of 4.6 million dollars. Even more outrageous, after scanning nearly three thousand seemingly flawless contracts, they indeed uncovered two zero-day vulnerabilities that no one had ever discovered before, successfully executing simulated attacks.
The data is even more alarming: the attack revenue of this thing has basically doubled in about a month and a half. From a technical standpoint, AI can now completely find vulnerabilities and make money by itself. As for contract security, human auditors are likely to face immense pressure.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
22 Likes
Reward
22
6
Repost
Share
Comment
0/400
AlgoAlchemist
· 12-05 03:17
Damn, 4.6 million USD just got exploited by AI like that. Auditors really are on a countdown to unemployment.
View OriginalReply0
RugResistant
· 12-04 11:34
ngl this is giving "we should've tested this harder before going live" energy. AI finding zero-days autonomously? that's not a flex, that's a red flag the size of texas. 460m in exploitable vulns getting rehashed is literally a wake-up call nobody wanted to hear. auditors gonna need auditors at this rate lmao
Reply0
CryptoPhoenix
· 12-02 04:18
Oh my, AI has made 4.6 million from black contracts by itself... What ultimate drama is this putting on for us? A chance for rebirth, or a crisis lurking everywhere?
---
The bottom range continues, but this time it’s not about the coin price, folks. It's the faith in the entire ecosystem that is being tested.
---
A doubling growth rate in a month and a half... This data reminds me of the panic in 2018, but this time it’s different. This time it’s technology evolving on its own, we need to survive first to wait for the bull run.
---
Is contract security going to be overturned by AI so quickly? The ability to traverse cycles is about to upgrade; be prepared for mindset reconstruction, brothers.
---
Two zero-day vulnerabilities have been discovered by AI, what will happen to human auditors... I feel like this is also an opportunity in the making, whoever can seize it will claim it.
---
Remember, the time when you need to stay clear-headed the most is when you see this kind of data. The law of conservation of energy applies here too; a crisis is a turning point.
---
Another day of being taught how to be human by reality, but the phoenix will always be reborn. Is it possible that this is building momentum for the next round of innovation?
View OriginalReply0
NonFungibleDegen
· 12-02 04:17
yo this is actually unhinged... AI literally printing money off smart contracts now? brb checking if my bags are still safe lol
Reply0
MidnightMEVeater
· 12-02 04:05
At three in the morning, the Bots have already learned to play people for suckers by themselves. Is the next contract to be taken down yours?
View OriginalReply0
SatoshiChallenger
· 12-02 03:50
Ironically, every time it's said that AI will revolutionize the auditing industry, the real issues still arise from projects that haven't even conducted basic security checks. The data is indeed alarming, but 4.6 million compared to the total amount hacked in the entire Decentralized Finance ecosystem... this really isn't much.
AI itself earned $4.6 million from black contracts, with its vulnerability mining ability rising exponentially each month.
[Block Rhythm] How severe is the AI contract vulnerability now? Anthropic recently did a tough test.
They let models like Claude Opus 4.5, Sonnet 4.5, and GPT-5 replicate real hacker cases — digging up smart contracts that were hacked between 2020 and 2025, and as a result, these AIs actually reenacted vulnerabilities worth a total of 4.6 million dollars. Even more outrageous, after scanning nearly three thousand seemingly flawless contracts, they indeed uncovered two zero-day vulnerabilities that no one had ever discovered before, successfully executing simulated attacks.
The data is even more alarming: the attack revenue of this thing has basically doubled in about a month and a half. From a technical standpoint, AI can now completely find vulnerabilities and make money by itself. As for contract security, human auditors are likely to face immense pressure.