arXiv:2606.07113v1 Announce Type: new Abstract: Large language models are rapidly becoming infrastructural components in high-stakes institutional settings, including public administration, legal reasoning, and healthcare, where opacity is not merely inconvenient but institutionally and legally untenable. Existing approaches to explainability are predominantly post-hoc, offering unstable, non-contestable accounts that have no formal relationship to the reasoning process that produced the output. We argue that the problem is not the absence of explanation but the absence of structured reasoning
The increasing deployment of large language models in critical sectors is revealing the limitations of current opaque 'black box' AI, making the demand for transparent and contestable AI paramount.
This shift from post-hoc explanation to integrated explainability ('glassbox AI') addresses institutional, legal, and ethical imperatives for AI governance, particularly in high-stakes applications.
The focus moves from merely explaining AI outputs after the fact to designing AI with inherent, formal, and verifiable reasoning processes, fundamentally altering AI development paradigms.
- · AI governance frameworks
- · High-stakes industries (e.g., legal, healthcare)
- · Explainable AI research teams
- · Regulatory bodies
- · Developers of purely opaque AI systems
- · Post-hoc explainability tool providers
- · Organizations relying on non-transparent AI without accountability
Increased trust and adoption of AI in previously hesitant sectors due to improved accountability.
New regulatory standards and certifications for 'glassbox' AI, potentially creating a competitive advantage for early adopters.
A potential re-evaluation of liability models for AI systems, shifting responsibility towards the design and formal reasoning processes rather than just output.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI