US technology executives are raising alarms about Chinese artificial intelligence models as Washington considers possible sanctions. The warning has gained attention after concerns about model safety, national security and the speed of competition between US and Chinese AI developers.
A central focus is DeepSeek, which was reported to show a 94% jailbreak success rate in testing. That figure has intensified debate over whether some AI systems can be too easily pushed past built-in safeguards, creating risks around misuse and harmful outputs.
Leaders at Anthropic and OpenAI have pointed to broader security issues tied to Chinese AI models. Their comments suggest the dispute is no longer only about commercial rivalry, but also about how governments may respond when advanced models are seen as a strategic and security concern.
With Washington now weighing sanctions, the issue is becoming part of a wider US-China technology confrontation. The latest warnings show how AI policy, model safety and geopolitical competition are increasingly colliding as both countries push for leadership in the field.
