A key executive in charge of Microsoft's (MSFT) artificial intelligence products said the security controversy surrounding Anthropic's AI model "Mythos" would prove to be more of an opportunity than a threat.
Marco Casalaina (pictured), Microsoft's vice president of products and core AI, met with Seoul Economic Daily at the company's Mountain View campus in California on Thursday and said, "Anthropic's Mythos is like a double-edged sword." He added, "It could be a threat, but at the same time it can be a means to block threats, so it will actually be an opportunity."
Mythos is a model that surpasses "Claude Opus 4.7," the top version of Anthropic's existing AI models. It is rated as the most powerful ever in its ability to identify cybersecurity vulnerabilities that have remained buried for decades, but it becomes equally dangerous if exploited by hackers. The AI industry has warned that if Mythos falls into hackers' hands, it could cause computer network paralysis, and Anthropic, considering the potential fallout, distributed the preview (trial) model only to major big tech and financial firms on a limited basis.
"We are cooperating with Anthropic to use Mythos," Casalaina said. "Our job is to find potential threats, warn companies and governments, and block security vulnerabilities." He continued, "Microsoft has several security tools such as Security Copilot, Purview, and Defender, and among them, Defender can filter out Shadow AI." Shadow AI refers to employees personally using external models rather than AI models officially approved within the company. It is cited as a risk factor because the company cannot maintain control when an employee enters confidential customer information or internal strategy documents into an external model.
"We must define what AI agent (autonomous) models should not do and create an environment where humans can intervene," Casalaina said.
