2 weeks ago
Apple
Apple's operating systems are known for their security, particularly compared to their rivals successful mobile and computing. Now, information researchers from a Palo Alto-based institution called Calif claim they were capable to breach macOS aft designing a privilege escalation utilization pinch thief from Anthropic's Claude Mythos Preview. As The Wall Street Journal reports, the utilization could beryllium utilized to entree parts of nan MacBook that should beryllium inaccessible and, thus, allows nan attacker to return power of a Mac computer.
The researchers worked pinch Mythos to place nan vulnerabilities and to thief them pinch nan exploit's development. Mythos Preview was capable to place nan bugs quickly, because they belonged to known classes. Human expertise was still basal to beryllium capable to creation nan exploit, but this shows precocious AI systems could unearth bugs and onslaught paths that were antecedently chartless and could beryllium utilized by bad actors to shape information breaches.
Apple is taking nan researchers' findings seriously, and told The Journal: "Security is our apical priority, and we return reports of imaginable vulnerabilities very seriously." In fact, nan researchers already met pinch nan institution astatine Apple Park successful Cupertino to talk what they're calling nan "first nationalist macOS kernel representation corruption utilization connected M5 silicon." If nan specifications they shared sounded vague, it's because they're readying to merchandise nan afloat method specifications of their findings aft Apple fixes nan vulnerabilities and onslaught path.
Anthropic uses Claude Mythos Preview for Project Glasswing, nan inaugural it launched successful April to forestall AI cyberattacks pinch AI. Glasswing participants, including Amazon Web Services, Apple, Cisco, CrowdStrike, Google, JPMorganChase, nan Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks, can use Mythos to fortify nan information of their ain projects. Mozilla, for instance, antecedently announced that it found and patched 271 vulnerabilities successful its latest merchandise of nan Firefox browser pinch thief from Mythos.
Just a fewer days ago, OpenAI introduced its ain cybersecurity inaugural successful consequence to Glasswing and Mythos. OpenAI's Daybreak uses its various AI models, including its specialized information supplier Codex. It was designed astir nan premise that cyber defense should beryllium built into package from nan opening and not conscionable revolve astir uncovering and fixing vulnerabilities.
