Trend Open models empowering less sophisticated malicious actors

Summary

The report expects that easier access to AI, especially open-source models, could empower less sophisticated malicious actors and raise security risks. AI misuse is already seen as enabling threat actors in the cyber domain and lowering barriers for new entrants. Generative AI can serve as an enabler of crimes such as fraud and the production of child sexual abuse material. Organised crime groups could use AI to become more effective and harder to detect, and because frontier AI might be used to design bioweapons, the risk connects to the acute risk of chemical, biological, radiological or nuclear attack. AI is also vulnerable to data poisoning, in which malicious actors insert harmful data to skew its outputs.