Publications |

The Consent Vacuum: When Agents Interact Without Humans

Tue, 07 Jul 2026 00:00:00 +0000

Existing consent frameworks assume a human who authorises an action in a context they can see. This paper asks what happens to a user’s authorisation once their AI agent stops talking to humans and interacts primarily with other agents. Using the early-2026 Clawdbot/Moltbook incidents as a stress test, it argues that consent degrades as an operative governance mechanism once it travels through open-ended agent-to-agent chains, a condition the paper calls the consent vacuum: not the absence of consent, but the structural erosion of the conditions that make it meaningful.

The paper contributes a taxonomy of six consent-failure types; a distinction between visibility, auditability and intervenability as oversight capacities relevant to Article 14 of the EU AI Act; and a mapping onto the GDPR, the AI Act and the Digital Services Act, with the DSA systemic-risk regime offered as a structural template for governing emergent agent ecosystems.

IAIL 2026 was held in Brussels on 7 July 2026. Co-authored with Osman Gazi Güçlütürk (Galatasaray University, Faculty of Law).

Research area: .

Consent Chain Degradation in Embodied Multi-Agent Systems: Bridging the Gap Between AI Agent Governance and Robot Ethics

Mon, 01 Jun 2026 00:00:00 +0000

Introduces consent chain degradation: the way the specificity, validity and scope of human consent weaken as authority passes through chains of delegation among embodied agents. The paper proposes CoRVE, a three-layer runtime architecture for consent scope modelling, delegation tracking and physical irreversibility assessment, illustrates the problem with healthcare, domestic and industrial scenarios (including a worked numerical example), and maps the gaps left by the EU AI Act, the GDPR, the Machinery Regulation and the revised Product Liability Directive.

Research area: , .

When Does Adaptive Guidance Help? Belief-Aware Privileged Distillation for Autonomous Driving Under Partial Observability

Mon, 01 Jun 2026 00:00:00 +0000

Extends Guided Soft Actor-Critic with Belief-Aware GSAC, which scales the privileged-distillation coefficient by ensemble disagreement in a driving POMDP. Across three occlusion levels, adaptive guidance helps under mild and moderate occlusion; under severe occlusion the coefficient collapses to its minimum early (“observability blindness”) and a simple linear decay schedule performs best, suggesting that much of the benefit comes from scheduling rather than adaptivity.

Research area: .

TR-TEB: Turkish Text Embedding Benchmark

Fri, 01 May 2026 00:00:00 +0000

Presents TR-TEB, a standardized and reproducible benchmark for Turkish text embeddings. It covers five task categories (classification, pair classification, clustering, retrieval and semantic textual similarity), combines 14 open-source resources, 26 translated datasets and 7 new Turkish-specific datasets, and evaluates 45 open-source embedding models.

Research area: .

From Chat to Agent: Why Consent Breaks When AI Acts on Your Behalf

Wed, 01 Apr 2026 00:00:00 +0000

A position paper on why consent mechanisms designed for conversational systems fail once AI moves from answering to acting on a user’s behalf. It identifies design tensions in agentic consent and offers provocations for rethinking user control.

Research area: .

Advancing geothermal reservoir temperature prediction in the Western Anatolia (Türkiye) using synthetic hydrogeochemical data from rank–Gaussian copulas