Steno Emphasizes Data Quality Risks in AI Training for LegalTech

According to a recent LinkedIn post from Steno, the company is drawing attention to risks it sees in current AI training practices. The post highlights commentary by Steno’s Senior Director of AI Products, Dan Ivtsan, in InformationWeek, focusing on the dangers of training models on AI-generated content.

The LinkedIn post suggests that relying heavily on synthetic data may preserve fluency while eroding factual accuracy, a degradation that standard benchmarks may fail to detect. Ivtsan’s view, as summarized in the post, is that synthetic data should augment but not replace real-world data.

For investors, this positioning indicates Steno is emphasizing data quality and model reliability, areas that could be critical differentiators in the competitive LegalTech and AI markets. By aligning with a more conservative, accuracy-focused stance on AI training, Steno may be targeting clients who prioritize trustworthy outputs in legally sensitive workflows.

This focus could support pricing power and client retention if enterprises increasingly scrutinize AI reliability and compliance risk. It may also signal future product development centered on monitoring, evaluation, and guardrails for AI systems in legal applications, potentially expanding Steno’s addressable market within #LegalTech and #LegalAI.