Скачать или смотреть ICRC curbs autonomous arms at AI For Good 2025 treaty push, digital emblem, drone UXO detection

Philippe Stoll, who oversees the humanitarian impacts of armed-conflict technologies at the International Committee of the Red Cross (ICRC), uses the AI for Good 2025 #aiforgood Summit in Geneva to demonstrate how machine‐learning, computer vision and decision-support tools already shape the battlespace and to argue that the same tools must be constrained by clear legal limits if they are ever to serve humanity rather than harm it. The ICRC’s exhibit walks visitors through battlefield scenarios where algorithmic target-recognition, sensor fusion and predictive analytics can either protect civilians or magnify risk, setting the stage for a broader discussion about “digital humanitarian law.” https://www.icrc.org/en/digital-emblem

He stresses a looming treaty push: the organization wants states to adopt, by 2026, a binding prohibition on autonomous weapons that select and engage human targets without meaningful human control. Stoll frames this as an extension of existing norms—especially predictability, traceability and command responsibility—that should accompany every new weapons technology. The urgency is reinforced by the May 12 2025 UN General Assembly debate on regulating “killer robots,” where consensus proved elusive despite mounting battlefield deployments.

Central to the ICRC argument is the Geneva Conventions’ distinction and proportionality test. Any AI-enabled fire-control or surveillance stack must reliably tell combatant from civilian and weigh collateral damage in real time. Stoll notes exploratory conversations with geospatial-intelligence and satellite-imagery providers who might supply independent data layers to audit targeting decisions—an example of technology serving legal oversight rather than replacing it.

Past precedents give the initiative weight. The Red Cross shepherded bans on blinding lasers in the 1990s and helped forge treaties on anti-personnel landmines and cluster munitions, proving that early diplomatic engagement can head off harmful deployments. Today’s agenda simply extends that pattern to neural-network-driven weapon systems, cyberspace infrastructure and fully automated command-and-control nodes, emphasising that states, not private vendors, carry the primary duty to codify the law.

While ICRC cannot regulate industry directly, technologists from defence primes to start-ups routinely seek its guidance. The organisation shares a public AI policy that spells out red lines—such as the need for explainability layers, bias-auditing pipelines and human-in-the-loop overrides at every lethal decision node—so companies can hard-code humanitarian constraints into software builds before pilot projects reach theatre deployment.

Internally, the Red Cross experiments with AI only when models are transparent and datasets ethically sourced. Logistics teams collaborate with EPFL and ETH Zurich on interpretable optimisation algorithms that allocate food, water and medical supplies without systematic discrimination, protecting vulnerable groups from inadvertent exclusion caused by skewed training data or opaque clustering routines.

A more kinetic research stream pairs ICRC with Waseda University in Tokyo, where edge-ML drones equipped with thermal-infrared cameras and multi-spectral sensors map unexploded-ordnance fields. The prototype can already flag metallic signatures on flat terrain and feed georeferenced heatmaps into humanitarian-demining workflows, though dense vegetation and mountainous topography still challenge the model and underline the need for rigorous field testing before operational deployment.

Finally, Stoll introduces the “digital emblem,” a cryptographically signed marker that labels ICRC servers and networks as protected humanitarian infrastructure. Much like a Red Cross flag on a field hospital, the emblem does not block cyber-attacks outright, but it creates a discoverable standard so that hostile actors cannot plead ignorance, and it places any deliberate breach squarely within the realm of war-crimes-prosecution, extending international humanitarian law into the digital domain to ensure future AI systems respect not only physical lives but also data integrity and humanitarian compliance.

https://www.reuters.com/sustainabilit...
https://timesofindia.indiatimes.com/w...

My full AI For Good 2025 video coverage is here:    • AI For Good 2025 video coverage  

This video was filmed using the DJI Pocket 3 ($669 at https://amzn.to/4aMpKIC using the dual wireless DJI Mic 2 microphones with the DJI lapel microphone https://amzn.to/3XIj3l8 )

Check out my video with Daylight Computer about their revolutionary Sunlight Readable Transflective LCD Display for Healthy Learning:    • Daylight Computer CEO interview: Healthy C...