Leistungsstörungen durch künstliche Intelligenz

Christian van Amelsvoort examines liability for contractual non-performance resulting from the use of artificial intelligence. To this end, he analyzes performance risk, counter-performance risk, and secondary liability risk, shedding light on their allocation from the perspective of the general law of obligations. The study also takes into account the relevant EU legislative initiatives, namely the AI Act, the withdrawn Proposal for an AI Liability Directive, and the preceding European Parliament proposal for an AI Liability Regulation. Particular attention is paid to strict attribution and evidentiary issues. The first part serves as a kind of general section, defining the autonomy risk of artificial intelligence as the factual subject matter of the inquiry and addressing overarching questions of attribution. It examines not only potential general limits to attribution but also cross-cutting grounds for attributing conduct or outcomes, including expected standards of care in the use of artificial intelligence and approaches to AI-specific strict liability. Of central importance, however, is the guarantee content of a simple promise of performance, which, drawing on the foundational work of Heck, may be described as a standard guarantee (Normalgarantie). The second part analyzes the individual levels of risk allocation on the basis of the relevant legal provisions. Throughout, attention is also devoted to possible evidentiary difficulties, for which broader solutions are discussed at the end, ranging from the reversal of the burden of proof and prima facie evidence to secondary disclosure obligations and a right to information grounded in the principle of good faith. The third part examines, in light of the statutory model developed in the preceding analysis and the specific prohibitions on contractual clauses, the extent to which risk allocation may be modified through standard terms and conditions. The study concludes with a set of summary theses and finds that there is no reason to fear a serious liability gap within the general law of contract.

Vorwort
Einleitung
A. Der Gegenstand der Untersuchung
B. Ausgeblendete Fragestellungen
C. Gang der Untersuchung
Erster Teil: Annäherung an den Untersuchungsgegenstand
A. Der Begriff Autonomierisiko
B. Generelle Grenzen der Risikozurechnung?
C. Generelle Gründe der Risikozurechnung
D. Ergebnis
Zweiter Teil: Gesetzliche Verteilung des Autonomierisikos
A. Das Autonomierisiko als Leistungsrisiko
B. Das Autonomierisiko als Gegenleistungsrisiko
C. Das Autonomierisiko als Sekundärhaftungsrisiko
D. Gesamtwürdigung der Beweislastverteilung
Dritter Teil: Gestaltbarkeit der Risikoverteilung
A. Verantwortlichkeit für das Autonomierisiko
B. Das Autonomierisiko als Leistungsrisiko
C. Das Autonomierisiko als Gegenleistungsrisiko
D. Das Autonomierisiko als Sekundärhaftungsrisiko
E. Beweislast für das Autonomierisiko
F. Ergebnis
Schlussthesen