The advent of AI tech may kick off an arms race in cybersecurity, but in the end what it’s most likely to do is harden targets until they are ...
Last week, many media outlets stirred up a panic about AI and cybersecurity. And the implication of the headlines was the exact opposite of the reality, which is becoming normal now.
The internet is going to be destroyed by AI. Why? Because a security vulnerability in the PHP language, used to code many websites, was discovered. The vulnerability was leaked to the public before security updates and patches could be widely circulated, allowing some hackers to attack websites.
What was often buried in these stories was the fact that it was AI technology that discovered the existing vulnerability, and AI technology that was being used to help fix this. The immediate opportunity for hackers was due not to the advent of “rogue AI” but to the normal human problem of bad operational security in human institutions.
The whole import of the incident was that AI is currently helping us to patch and fix the “technical debt” that institutions have accumulated for years — the little loopholes and exploits and unlocked doors are being discovered, usually by the good guys, who are the security experts with access to the most advanced models and the most powerful computers.
There are lots of problems with AI in education, in the way chatbots may talk to children, or in the way it provides shortcuts to human intellection and creativity. The advent of AI technology may kick off an arms race in cybersecurity, but in the end what it is most likely to do is harden targets until they are truly invulnerable.