Responsible AI Use: Awareness, Risks, and Good Practices

The project reduces AI-related existential risk by addressing failure modes that emerge from widespread, uncritical adoption of increasingly capable AI systems. In particular, it targets risks associated with epistemic degradation (misinformation propagation), automation bias, and the integration of unreliable AI-generated outputs into critical software and decision-making pipelines.

By focusing on software professionals and technically literate audiences, the project intervenes at a key leverage point: developers and early adopters who operationalize AI systems at scale. Through publications and events, it promotes practices such as output verification, adversarial thinking about model behavior, secure development workflows, and maintaining human oversight in high-stakes contexts.

This contributes to mitigating systemic risks in several ways: reducing the likelihood of cascading failures caused by erroneous AI-generated code or data, limiting the amplification of false or misleading information, and increasing collective awareness of model limitations and misalignment risks. It also helps counter overreliance on AI systems in domains where robustness and correctness are critical.

In the long term, the project supports the development of a more safety-conscious ecosystem by shaping norms around responsible AI use. These shifts can improve the reliability of human-AI interaction, slow down unsafe deployment patterns, and create stronger demand for aligned, interpretable, and verifiable AI systems.