tech
The biggest barriers to AI-powered cybercrime are disappearing
Three trends are opening the door to more AI-assisted attacks.
TL;DR
- Barriers like restricted access, high costs, and lack of incentive for cybercriminals to change tactics are falling for AI adoption.
- Open-weight AI models are approaching the capabilities of commercial models, offering more options to attackers.
- Underground marketplaces now offer jailbroken AI models, custom models, and AI-powered hacking services.
- Hackers are becoming more comfortable integrating AI into their existing workflows.
- Recent attacks, such as the JadePuffer incident, show AI agents automating tasks like writing exploit code and negotiating with victims.
- AI is empowering individual attackers with significantly more leverage.
- While open-source models exist, attackers still favor models like Claude and ChatGPT due to technical expertise and resource requirements for open-source alternatives.
- Defenders have a shrinking window to prepare for AI-accelerated attacks.