Chief Information Security Officers (CISOs) are sounding the alarm over DeepSeek, a Chinese AI model, warning of serious cybersecurity and data privacy risks. Anxiety is growing in security operations centres globally as organisations grapple with the AI’s rapid adoption and potential vulnerabilities.
Global scrutiny intensifies
Several countries have already restricted or banned DeepSeek. The U.S. Navy has prohibited its use, while the Czech Republic barred it from state administration citing cybersecurity concerns. Italy, Australia, and South Korea have introduced similar measures, highlighting fears that sensitive data could be accessed by foreign authorities.
Vulnerabilities in the AI
Security research indicates DeepSeek’s R1 model fails to block harmful prompts, leaving it open to manipulation and algorithmic “jailbreaking.” Experts warn this exposes organizations to potential cyberattacks, disinformation campaigns, and misuse in high-risk environments. The model’s safeguards lag behind other leading AI systems, raising alarm among CISOs.
Data privacy and national security risks
DeepSeek collects extensive user information, including prompts, responses, and device data, stored on servers in China. Security experts caution this creates a risk of data exposure to Chinese authorities and potential exploitation for surveillance or intelligence-gathering purposes.
Calls for regulatory action
In response, CISOs are urging stricter AI governance. U.S. lawmakers have introduced the “No Adversarial AI Act,” which would prevent federal agencies from using AI developed in adversarial nations like China. The legislation aims to shield critical networks from foreign influence and reinforce standards for AI security and accountability.
Balancing innovation and security
While AI technologies offer significant business and operational benefits, security chiefs emphasise the need for robust regulation to prevent misuse. Ensuring safe deployment, transparent data handling, and enforceable security standards is becoming a priority for governments and enterprises alike.
REFH – Newshub, 19 August 2025
Recent Comments