Anthropic reported to government and major private institutions that its AI model, Claude Mythos, found thousands of security vulnerabilities in widely used software.
The New York Times described it as an “AI-driven cyber threat”, but Wall Street Journal columnist Holman W. Jenkins Jr (홀먼 W. 젠킨스 주니어) took a different view, saying Mythos would also have a positive impact on security.
In a recent column, he stressed: “Mythos did not create vulnerabilities; it found them. The difference matters. Software security vulnerabilities already existed. Mythos found them before malicious actors did. It is a net gain to the world, not a net loss.”
He said capitalist incentives influenced Anthropic’s decision to disclose what it found about Mythos.
He said it was the result of litigation risk, regulatory pressure and opportunities the company could seize in the future security market coming together. “If the same technology were in the hands of a Chinese company, the result would have been different. The vulnerabilities found would have been reported to the Communist Party first. At this very moment, hostile governments and criminal organisations are also looking for the same vulnerabilities,” he said.
He said Mythos would also be positive for Anthropic commercially.
“Anthropic can turn the vulnerabilities found by Mythos into a multi-billion-dollar security business. Social contribution and business interests point in the same direction. Anthropic CEO Dario Amodei (다리오 아모데이) and the company are portrayed as altruistic heroes, but in reality this is a case where capitalist incentives worked properly,” he said. “Broadening the perspective, Mythos can also be seen as a case that shows AI’s potential to make better judgment cheaper.”
“If governments and companies routinely use AI for decision-making, choices based on unfounded optimism or biased judgment could decline. AI does not look away from uncomfortable truths that humans may want to avoid. Mythos showed that potential first,” he said.
Click Here For The Original Source
