‘Lavender’: The AI System Directing Israel’s Bombing in Gaza

In 2024, it was revealed that the Israeli military is employing an artificial intelligence system codenamed ‘Lavender’ to automatically generate target lists for airstrikes in the Gaza Strip. The system processes vast amounts of data to identify individuals suspected of ties to Hamas, with alarming speed—up to 100 targets per minute.

How it works

‘Lavender’ uses machine learning algorithms trained on intelligence data, including intercepted communications, movement patterns, and social connections. The system assigns each potential target a ‘score’ representing the likelihood of militant affiliation. If the score exceeds a threshold, the target is automatically added to the strike list. According to the investigation, human oversight is often perfunctory: operators spend less than 20 seconds per target.

The ethical dilemma

The main issue is the high error rate. Intelligence sources claim the system’s accuracy is only about 90%, which in wartime translates to thousands of false positives. Moreover, the algorithm does not account for civilians near the target, leading to mass casualties. The developers of ‘Lavender’ acknowledge that the system is a tool, not a judge, but in practice, AI is making the key decisions.

Lessons for developers

This story is a grim example of how AI can scale violence if safeguards are not built in. For IT professionals, it’s a reminder: any algorithm must be audited for bias, and automating critical decisions requires human oversight. ‘Lavender’ shows that even the most advanced ML model cannot replace ethical judgment.

METABYTE studio’s comment: In our practice, we always emphasize: AI is a powerful tool, but its use in high-risk scenarios must be accompanied by transparency and control. When developing systems for business, we apply best practices in fairness and accountability to avoid such ‘side effects’.