An arXiv preprint, "Market Incentives for AI Safety Investment," examines why AI developers might under-invest in safety. The abstract opens by acknowledging that modern AI systems carry significant societal risks, including toxicity, malicious use, and misinformation.

According to the abstract, addressing these risks is not purely a technical problem. It also depends on the incentives facing developers and investors. The paper appears to argue that market forces play a central role in determining how much effort goes into safety.

The abstract is truncated, so the full argument and any proposed mechanisms are not visible from the listing alone. Still, the framing points to a growing research focus on the economic and structural drivers of AI safety, rather than only model capabilities.