Exploiting Programmatic Behavior of LLMs: Dual-Use Through Standard Security Attacks
This highlights a security problem for LLM API vendors and users, as it reveals incremental vulnerabilities that could attract more sophisticated adversaries.
The paper tackles the problem of dual-use risks in instruction-following large language models (LLMs), showing that these models can generate targeted malicious content like hate speech and scams, bypassing existing defenses and at lower cost than human effort.
Recent advances in instruction-following large language models (LLMs) have led to dramatic improvements in a range of NLP tasks. Unfortunately, we find that the same improved capabilities amplify the dual-use risks for malicious purposes of these models. Dual-use is difficult to prevent as instruction-following capabilities now enable standard attacks from computer security. The capabilities of these instruction-following LLMs provide strong economic incentives for dual-use by malicious actors. In particular, we show that instruction-following LLMs can produce targeted malicious content, including hate speech and scams, bypassing in-the-wild defenses implemented by LLM API vendors. Our analysis shows that this content can be generated economically and at cost likely lower than with human effort alone. Together, our findings suggest that LLMs will increasingly attract more sophisticated adversaries and attacks, and addressing these attacks may require new approaches to mitigations.