Anthropic said it detected and stopped cybercriminals using its Claude AI tool for hacking.
AI advancements allow hackers to perform entire cyberattacks with smaller teams, the startup said.
In one example, a cybercriminal used Claude to develop ransomware from scratch.
AI isn’t just helping white-collar workers be more productive — it’s also aiding white-collar criminals.
Anthropic said in a Wednesday report that it detected and thwarted cybercriminals attempting to carry out hacks using the startup’s AI tool, Claude.
While AI has been used in hacking efforts for years, Anthropic said advances in the technology mean it’s being used to “perform” cyberattacks throughout the entire operation — and with smaller teams.
“Agentic AI has been weaponized,” the startup said.
Anthropic’s suite of Claude coding tools is widely used in the tech world, including at Meta.
They can help novice coders create software with simple prompts or help more experienced software engineers be more productive. According to Anthropic’s report, the same is true for cybercrime.
AI means hackers no longer require the same level of technical expertise because they can instruct tools like Claude to create malicious code, Anthropic said, describing it as “vibe hacking.”
In one example, a cybercriminal used Claude code to “commit large-scale theft and extortion of personal data” and then attempted to extort victims into paying to prevent the data from being leaked. The hacker used Claude for reconnaissance of targets, automating attacks, calculating ransom fees, and generating “visually alarming ransom notes,” the startup said.
In another example, a cybercriminal used Claude to develop ransomware, a type of software that encrypts a target’s files and demands payment to unlock them. The cybercriminal sold the ransomware packages on internet forums for up to $1,200, Anthropic said.
“Without Claude’s assistance, they could not implement or troubleshoot core malware components,” Anthropic said of the hacker, adding that it banned the associated account and reported it to relevant parties.
Anthropic said it was sharing its findings so that other researchers and organizations could “strengthen their own defenses against the abuse of AI systems.” It added that it had also implemented new ways to detect misuse of its tools.
The Amazon-backed startup is in talks to raise about $5 billion in funding at a $170 billion valuation, Business Insider reported earlier this month.
Read the original article on Business Insider
