# Frontier Models Turn Up in Weapons Targeting

Lab threat reporting increasingly documents attempts to use general-purpose frontier models for lethal and military applications such as autonomous drone targeting, so expect recurring named weapons-misuse categories in safety reports, account-level enforcement as the first-line control, and pressure to treat model access as a dual-use export question rather than a content-moderation one.

- Conviction: 36 / 100 (forming)
- Horizon: Emerging (watchlist)
- Tracking since: 2026-09-15T00:00:00.000Z
- Last updated: 2026-09-15T06:50:06.768Z
- Canonical: https://polylog.news/ai/trends/frontier-models-in-weapons-targeting
- Publisher: Polylog
- Affected regions: United States, China, Russia / Ukraine

## Recent evidence

- [confirms] Anthropic says it disrupted Claude misuse across seven harm areas, including autonomous drone targeting (2026-09-15): Anthropic's September 10 misuse report lists autonomous drone targeting among seven harm areas it disrupted between December 2025 and August 2026. It is the first appearance of a lethal-targeting category in a mainstream frontier-lab threat report, distinct from the fraud and intrusion categories already tracked.