Off-By-1-Labs study finds AI fixes 26% of software vulnerabilities
Technology
A new study from Off-By-1-Labs found that artificial intelligence could only fix software vulnerabilities 26% of the time, out of more than 6,000 tries.
Even worse, over half the patches made by AI either didn't solve the problem or actually introduced new bugs.
Turns out, letting AI handle security fixes on its own isn't quite ready for prime time.
'FLAWED' on GitHub, AI needs oversight
While AI can help spot and sort out potential issues, it's not reliable enough to patch them without human help.
The researchers have shared their FLAWED tool on GitHub, but for now, real experts are still essential in keeping our software safe.