Palo Alto CEO warns AI will accelerate cyberattacks
What's the story
Nikesh Arora, the CEO of Palo Alto Networks, has warned that cyberattacks are likely to increase in frequency. He made the prediction in a post on X, saying, "Attacks will accelerate from here on." Six months ago, he had predicted that capabilities similar to Anthropic's Mythos would be widely available. Now, he says that this is already happening.
Attack trends
Cyberattacks now being witnessed in the wild
Arora pointed out that cyberattacks are now being witnessed in the wild.
These attacks exploit a critical vulnerability, use a harmless-looking social engineering trick, or chain together publicly available vulnerabilities in a "sniper strike."
He said incident responders have never been busier than they are today due to this rise in cyber threats.
Attack sophistication
Attacking has become easier than before
Arora clarified that cyberattacks aren't necessarily becoming more sophisticated. Instead, they're happening faster and by a larger number of people.
He explained that attack path discovery, which once required trial and error, is now trivial with the help of open-source LLMs (large language models).
This means more people are getting better at hacking as attacking has become easier than before.
Security measures
Supply chain security is crucial
Arora emphasized the importance of supply chain security, saying, "you are as secure as your supply chain."
He warned that insecure upstream or downstream players with access to a company's systems could be the next attack vector.
He also said marketplaces and supply chains are easier targets for cybercriminals.
Security advice
What can organizations do?
To combat the rising threat of cyberattacks, Arora has advised organizations to take certain steps.
These include testing their infrastructure for accessible vulnerabilities and misconfigurations, fixing their own code while demanding the same from vendors, and updating old or insecure infrastructure at the perimeter first.
He also stressed putting a security framework around new AI tools like coding tools since these models are trained on human code.