
⚠️⚠️ JUST IN: BlockSec retested EVMBench, the AI code test from OpenAI and Paradigm. They say bots did worse in real hacks than EVMBench claimed: 72% success at attacks and 45% at finding bugs on 120 cases.
This shows AI cannot replace human security experts yet. Watch for more tests with new bugs, and for teams to pair AI with human checks.
#AI #Security theblock.co/post/394553/ai…
English