AI targeting systems Lavender and Gospel violate international humanitarian law by design, academic analysis finds
By
Authors: Rainer Rehak, Taylor Kate Woodcock
Summary
This academic article analyzes three AI-based targeting systems used by the Israel Defense Forces in Gaza since 2023 — 'Lavender', 'Gospel', and 'Where's daddy?' — and evaluates their compliance with International Humanitarian Law (IHL). Drawing on expertise from computer science, science and technology studies, and international law, the authors describe the technical architecture of these systems, their reliance on US tech infrastructure (Microsoft, Amazon, Google), and their configuration for target production and civilian harm. The evaluation against IHL principles of distinction, precautions, and proportionality finds that the systems violate all three principles by design. Key findings include: (1) the systems show reckless disregard for lawful target selection, (2) they enable mass target generation without substantive legal justification, and (3) their probabilistic nature makes them akin to indiscriminate 'dumb bombs'. The authors argue that civilian casualties are not accidents but directly configured into the systems, and that the term 'targeted killing' serves as a discursive red herring to maintain an appearance of ethical conduct.
Source

Key quotes
· 5 pulledThe explicit and implicit configuration of the desired outcomes, including civilian harm, constitutes a reckless disregard for the selection of lawful targets, hence violating the principles of distinction and precautions under IHL.
The systems enable entirely new modes of operation by allowing an arbitrary number of targets, including high numbers of unintended engagements of civilians, to be generated per day with virtually no substantive and legal justification.
These AI-enabled systems are probabilistic in nature and, as a consequence, more akin to traditional large indiscriminate 'dumb bombs', hence violating all three key principles of the IHL framework when relied upon uncritically.
The masses of civilian casualties are not 'accidents' in an otherwise 'precise' system but that they have been directly configured into the systems.
The massively lethal use of such systems can be plausibly assumed to be deliberate. Yet, still describing their function as 'targeted killing' could furthermore be understood as discursive red herring for keeping up an appearance of ethical conduct.
You might also wanna read

Israel's AI Weapons Systems in Gaza Raise Humanitarian and Legal Concerns
The article examines Israel's development and use of AI-powered weapons and surveillance systems in Gaza, raising serious ethical, legal, an
thecairoreview.com·10mo agoIsrael identified 1,000 'real-time targets' per day over two years in Gaza, Lebanon using Elbit Systems AI
Palantir's AI Platform Used by U.S. Military to Track Gaza Aid Deliveries
Palantir Technologies' AI platform is being used by the U.S. military to track and coordinate humanitarian aid deliveries in Gaza through it

Project Maven: How AI transformed US military targeting and sparked Silicon Valley controversy
The article examines Project Maven, an AI-powered targeting system developed by the US military in partnership with Silicon Valley, which dr

The reality of AI warfare: How autonomous weapons have already arrived
The article discusses the current state of AI warfare, focusing on lethal autonomous weapons systems (LAWS). It highlights that while intern

Comments
Sign in to join the conversation.
No comments yet. Be the first.