Google puts new limits on AI agents that hunt threats and fix vulnerabilities

Goo­gle Cloud is ad­ding new con­trols to Co­de­Men­der, its AI tool for fin­ding and fi­xing soft­wa­re vul­ne­ra­bi­li­ties, allo­wing bu­si­nes­ses to iso­la­te at­tack si­mu­la­tions and tar­get code scans more pre­ci­sely. The up­da­tes let bu­si­nes­ses ve­rify vul­ne­ra­bi­li­ties and si­mu­la­te ex­ploits in an ope­ra­ting sys­tem-level sand­box or an iso­la­ted vir­tual ma­chi­ne. They also add human checks for local wri­tes and com­mand execu­tion to help keep tes­ting within de­fi­ned li­mits. Bu­si­nes­ses can scan an en­ti­re code re­po­si­tory, a spe­ci­fic … More → The post Goo­gle puts new li­mits on AI agents that hunt th­reats and fix vul­ne­ra­bi­li­ties ap­pea­red first on Help Net Se­cu­rity .