Governance-Driven Logging, Transparency, and Human Oversight

Identify which logging, transparency, and human-oversight affordances coding-agent architectures should expose to satisfy emerging governance and regulatory obligations such as the EU AI Act and the GPAI Code of Practice.

Background

The paper notes increasing external constraints from governance frameworks (e.g., EU AI Act, GPAI Code of Practice) and safety reports. While Claude Code’s deny-first permissions and append-only transcripts aid internal auditability, it is unclear what externally auditable interfaces are required.

Answering this question would inform compliance-oriented design of audit logs, policy controls, and oversight hooks within agent architectures.

References

Which logging, transparency, and human-oversight affordances coding-agent architectures should expose under that external constraint remains an open design question.

— Dive into Claude Code: The Design Space of Today's and Future AI Agent Systems  (2604.14228 - Liu et al., 14 Apr 2026) in Section 12.5 (Governance and Oversight at Scale)

RGE tracks that relation online without an end-to-end LLM verdict, leaving open what evidence interfaces future agent environments should expose to make closure, authorization, and surplus action monitorable.

— Beyond Suspicious Steps: Ontological Trust in Long-Horizon Agents  (2608.17718 - He et al., 18 Aug 2026) in Conclusion

The substantive question this raises---which the framework leaves open---is which reasons of which agents a system ought to track.

— Anticipatory Human Oversight of Agentic AI: A Philosophical Account  (2609.24242 - Baum et al., 21 Sep 2026) in Section 2.3, subsection “Meaningful Human Control and the Reasons Foundation”

Whether or not the impermissibility result under our framework also implies a violation of Article 14 AI Act must remain an open question. Article 14(3) AI Act demands oversight measures to be commensurate with the risks, level of autonomy, and context of use of the high-risk AI system. An answer can thus only be found on a case-by-case basis.

— Anticipatory Human Oversight of Agentic AI: A Philosophical Account  (2609.24242 - Baum et al., 21 Sep 2026) in Section 3.5, subsection “Three Architectural Desiderata”