Anthropic’s exposure of blue-chip security flaws is a seminal moment
Ten years ago, almost to the day, I used my inaugural Dialogue column to muse that ‘Apple needs its man from Mars.’ Said Martian was none other than Elon Musk, who told me he planned to retire to the Red Planet.
I urged him to wait and instead partner Tesla with Apple so the Cupertino giant could launch a car. The Apple Car has yet to arrive, but the intervening years have seen many greater innovations. Prime among them is AI – something Musk has warned will take all our jobs. He may be right. Yet the immediate concern is not our employment, but our cybersecurity.
This spring, Anthropic revealed that its latest AI system, Claude Mythos, can run riot over elite operating systems. It can discover, chain and weaponize software vulnerabilities at warp speed – far beyond anything humans can achieve. Its efficacy at identifying thousands of security holes triggered a meeting of minds.
The greatest menace is that of speed. The Claude case rewrites the rules of cyber conflict. Once, cybercriminals relied on hoodwinking humans into clicking malicious links. Now, that seems vaguely quaint. The advent of robotic superhackers means the entire lifecycle of an attack – the reconnaissance, exploitation and persistence – can now run as a continuous, autonomous process operating faster than humans can respond.
If the threat is huge in the West, it is monumental in emerging economies. I live in California. Given that the state is the home of AI, you would hope businesses here would be more cybersecure than they are. Incompetence – even denial – is commonplace just a few miles from Silicon Valley. Yet compared to my native India, the Americans look like geniuses. Communications in the world’s most populous nation – from businesses to government – run mostly on WhatsApp. Thus commercially – even perhaps militarily – sensitive information is being traded on a platform owned by Big Tech.
The increasing power of a handful of tech giants over the world economy is the critical shift of the modern age. Major governments have noticed and are uneasy; note the growing tension between the US Department of Defense and Anthropic. The Pentagon is right to worry. AI is no longer merely a commercial technology. It now sits at the crucible of national security. The Valley retains autonomy over its epoch-defining products. In them, it holds the levers to shape the global balance of cyberpower.
Back in 2016, when I began this column, nothing worthy of the name ‘artificial intelligence’ existed. In my ten years as Dialogue’s technology columnist I have tried, as best a mere-mortal can, to keep pace with and analyze the great opportunities – and profound threats – machines can deliver. I remain hopeful that the gains will exceed the losses: cures to global diseases are possible with the might of today’s computational power.
Yet, as I depart after a fascinating decade, I urge Dialogue readers to remain inventive, industrious and ethical. Above all, they must be vigilant. It was once the command of armadas and arsenals that delivered power to nations. Now it is the command of AI. The hand that rocks that cradle will rule the world.
Vivek Wadhwa is author of The Immigrant Exodus: Why America Is Losing the Global Race to Capture Entrepreneurial Talent
