AI "Agents" extend their deployers – references concerning digital legal accountability

There have been some disastrous "decisions" made or floated in the last few days, so a cybersecurity expert who knew me from my talks asked me for my related papers, and I've decided to put them in one (editable) place. Because I think cybersecurity is going to be bigger than AI in the next 7 years. 

Academic papers, chronologically:

  • J. J. Bryson, D. L. Martin, S. A. McIlraith and L. A. Stein, "Toward behavioral intelligence in the Semantic Web," IEEE Computer, vol. 35, no. 11, pp. 48-54, Nov. 2002, doi: 10.1109/MC.2002.1046974.  https://ieeexplore.ieee.org/document/1046974
  • Bryson, J.J., Diamantis, M.E. & Grant, T.D. Of, for, and by the people: the legal lacuna of synthetic persons. Artif Intell Law 25, 273–291 (2017). https://doi.org/10.1007/s10506-017-9214-9 fully open access
    • The first time we said "An AI legal agent would be the ultimate shell company." We had honestly expected this would push back on the overextension of legal agency for corporations more generally, we couldn't believe the terrible idea of AI legal agency would last for long (though we did have international lawyers ready to challenge any EP law about this.)
  • Evans, K.D., Robbins, S.A. and Bryson, J.J. (2025), Do We Collaborate With What We Design?. Top. Cogn. Sci., 17: 392-411. https://doi.org/10.1111/tops.12682 fully open access
    • tl;dr no. AI is capital, not labour. It is not a peer. Its priorities are chosen by a weird mix of its developers, its deployers, any hackers who've got in, and if you're very lucky, a bit of the user's intent.
  • Not me but an outstanding book: "The Problem of Personhood: Giving Rights to Trees, Corporations and Robots" by Lisa Siraganian https://www.versobooks.com/products/3061-the-problem-of-personhood
Blogposts:
  • Legal personality for artefacts breaks justice and democracy AI is not countable, not even AI agents. They can always be cloned (that is, arbitrarily multiplied,) deleted, or entirely rewritten. Or hacked.
  • No One Should Trust AI (published by the UN University). It is not the kind of thing you can form a relationship with (see above). It is however possible and desirable to engineer Artificial Intelligence for accountability.
See also the rest of my recent publications; many are relevant, and the policy and security links here on this blog.

Comments

Post a Comment