OpenAI Publishes National Security Principles For Government Partnerships

From the new National Security Principles. OpenAI does not support models being used for: - Mass domestic surveillance - High-stakes decisions without appropriate human judgment and accountability - Use of force without appropriate human judgment and accountability - Evading legal obligations, oversight, or accountability 'These principles do not categorically exclude the use of OpenAl technology to conduct intelligence operations, investigations, or offensive and defensive military operations. We do not think the right line is a categorical distinction between offense and defense since, in practice, a capability or operation may be characterized as strategically or tactically offensive or defensive depending on its purpose, context, and intended effects. The relevant question, therefore, is whether those uses are consistent with the principles and remain within the limits outlined above.' https://openai.com/index/government-national-security-partnerships/