Signed in as:
filler@godaddy.com
Responsible AI Governance
AI and the Military
For the past six years, INHR has been convening international dialogues about safety of AI military systems with retired generals, diplomats and current tech professionals. Our conclusion, after discussion with US, European, Chinese and Indian experts: There is no one-size fits all answer on how to govern sensitive AI applications. Bans are impractical, the UN is not suited to regulate, and limits on national law prevent investigation or remediation when an AI incident occurs. INHR experts have developed a voluntary Code of Conduct for mitigating the risks of AI weapons, a guide to military AI testing and evaluation, and recommendations to make AI's use in biosecurity safer from terrorist abuse. We offer training on these and other practical approaches to Responsible AI for companies and countries alike.
INHR Proposes Int'l Governance Solutions for biosecurity and AI accidents at REAIM 2024
INHR presented its ideas to make military AI safer and more responsible to a packed side event at REAIM 2024 in Seoul September 10. The film above shows our ideas for international AI governance improvements. Gen. Jack Shanahan (US), Gen RS Panwar (India), Dr. Iben Yde (Denmark) and Dr. Barry O'Sullivan (Ireland) led a discussion of AI military risk mitigation. They discussed maintaining human control over nuclear command & control, whether decision-support systems and AI targeting comply with IHL, and how to deal with online learning and generative AI when designing AI military governance.
How dangerous are the risks of AI and Biosecurity Misuse?
Thanks to Founders Pledge, INHR's multilateral dialogue on AI military safety began in May 2024 to explore the risk that terrorists might use AI tools to create biological weapons. Governments and the international community lack the capacity and institutions to prevent and respond to AI x bio risks. Companies devise guardrails for Large Language Models and Biological Design Tools to prevent misuse, but malign actors "jail break" and find loopholes in these safety measures. INHR hopes its multistakeholder dialogue helps the international community stay one step ahead of malign actors by building capacity and sharing model safety practices to mitigate these risks.
AI and Human Rights
At the AAAI-23 conference in Washington DC, Eric Richardson explained for computer scientists and data professionals why diplomacy is important to the future of AI. He provided an overview of UN Human Rights Council resolutions touching on AI and UNESCO's AI Ethics Rrinciples (see blog below). As a human rights NGO, we know how important it is to protect data privacy, fight bias, and ensure that AI is a force for good in human rights. We promote AI that is responsible, ethical and inclusive - especially for women, minorities and the developing world. But as technology savvy professionals, we are not afraid of AI and know that AI applications can also help human rights defenders document rights violations around the globe. Informing international legal and human rights principles with technical considerations and a deeper understanding of AI, as the UN Human Rights Council has done in resolutions 47/23 and 53/29, is a constructive way forward, we believe.
Copenhagen Conference on International Cooperation in AI-Enabled Weapons and Security
In August 2023, INHR co-hosted a Track II AI dialogue at Frederiksberg Castle in Copenhagen, with our sponsors Center for New American Security (CNAS) and the Royal Danish Defence College! We were pleased to meet with multi-domain experts from the U.S., Europe, China and India. The conference was a success, with productive discussion from all members, in-person and online. For more details see here our Model Practices Guide to Test and Evaluation of AI-enabled military systems!
AI and the UN
UN regulation of artificial intelligence has been scattershot to date, but we hope that the UN Secretary General's new expert group can help to bring coherence to the international playing field. When the UN focuses on AI as a way to get more donor funding, its expertise and recommendations for governance can be troubling. OHCHR's B-Tech, UNESCO's
INHR @ REAIM, AAAI-23 and LAWS Conferences
When serious discussion of AI governance occurs, INHR is there, including our attendance or presentation at recent events such as:
- the AAAI-23 Conference in Washington DC (Feb 2023)
- the REAIM Conference in the Hague (Feb 2023)
- the Luxembourg LAWS conference (May 2023)
- the UNIDIR Innovations Dialogue (June 2023), and
- the 53rd Human Rights Council discussion of new technologies (June/July 2023).
We host our next trilateral meeting with the Royal Danish Defense University in Copenhagen in August 2023.