Human Oversight of AI
TLL Global Management LLC, in partnership with Practera and King’s Business School (King’s College London), is researching human oversight of AI in the workplace.
The question at the centre of the work: can you show the stop?
What the research found
- Regulators require oversight in principle (EU AI Act Articles 14 and 26). Organisations need it in practice. Between the two sits a gap nobody measures.
- Article 14 asks for systems that can be stopped. Article 26 asks for real people with the authority to stop them. The gap opens where a person is named but not empowered.
- Being named as an overseer is not enough. The person needs authority, information, time and support to veto a decision.
- Proof of oversight is becoming a precondition of insurance cover. Insurers price AI risk now, asking for autonomy levels, human checkpoints and rollback procedures.
- Oversight covers access as well as output. Who holds the keys, who can revoke them, and who acts the moment someone leaves.
The final report, Human Oversight of AI: Beyond Compliance, will be available for download here shortly.
In partnership with
Past Projects
1 projectConsensus Document
2024At our first Global Summit — “The Intersection of Climate, AI & Humanity” — Tech Leadership Lab brought together thought leaders, innovators, policymakers and activists. The summit produced United for Tomorrow: Catalyzing Climate Solutions at the Nexus of AI and Humanity, a 26-page consensus on the findings, opportunities, challenges and unanswered questions that emerged.
Download the Consensus Document