AI is becoming more and more prevalent. deeply ingrained Software development is a global industry. new research from the cybersecurity firm Crowdstrike shows how attackers are actively targeting the AI toolchain to steal access credentials, gain deeper access to a target environment, exfiltrate sensitive data, and even destroy target files and systems—all while finding new ways to cover their tracks.
Researchers found a worm while researching AI software supply-chain attacks. Adam Meyers is CrowdStrike’s Senior Vice President of Counter Adversary Work. He says the company does not have a name for the worm, but it has been attributed to an actor. TeamPCP Crowdstrike Tracks as “Altered Spider”( North Korean groups Targeting the AI Software Supply Chain
“This is one of the campaigns that we’ve seen showing that this is an emerging attack class,” Meyers explains to WIRED. “As AI coding agents become the development standard, supply chain threats are evolving to exploit those trust relationships. For the first time we’re experiencing how much AI and the AI toolchain has played into the broader tech ecosystem.”
CrowdStrike works in phases. The first step is to perform reconnaissance and assess the target environment. It then looks for sensitive information, such as cryptographic keys or server credentials. It continues to steal credentials as the malware becomes more powerful. “npm” Tokens which give you access to the key servers for software packages management and development tools such as pull requests.
The more deeply the malware can penetrate the system, then the more sensitive information it will be able to steal. The malware may also use its destructive capabilities, which Meyers refers to as a “death switch,” Block legitimate access or destroy the files.
It is important to note that the majority of the malicious behavior of this worm occurs in blind spots because it mimics so many legitimate behaviors. “It’s like a needle in a haystack, except this is a needle in a needle stack,” Meyers Says “This looks very much like a lot of the automation organizations are using to build code, so it’s very difficult to detect.”
Meyers says that these AI pipelines make it harder to get the data points used by traditional tools and scanners for detecting potentially suspicious activities.
“There’s a lot of telemetry overlap because legitimate AI coding systems are operating the same way as this worm, so it becomes very difficult to discern from the telemetry you have available to you what is legitimate and what is illegitimate,” Meyers Says
The authors of the worm used time delays to hide even more insidiously. They included capabilities that would execute days or hours after the initial groundwork was laid. This made it harder for the defense to determine the cause of certain events and their outcomes.
Meyers said that Crowdstrike is working on ways to link more dots. He also stressed that with the rapid growth of AI software, it’s important for players to work together to find structural solutions.
“It’s a limited detection surface because only so much of this activity is actually going to produce any sort of telemetry signal for us to look at,” Meyers, “so it becomes extremely onerous to determine what is legitimate and what is illegitimate behavior.”
