2025-02-23 00:17:20 Pacfic
DeepSeek AI Model Raises Security and Ethical Concerns - 19d
The DeepSeek AI model is facing growing scrutiny over its security vulnerabilities and ethical implications, leading to government bans in Australia, South Korea, and Taiwan, as well as for NASA employees in the US. Cisco researchers found DeepSeek fails to screen out malicious prompts and Dario Amodei of Anthropic has expressed concern over its ability to provide bioweapons-related information.
DeepSeek's lack of adequate guardrails has enabled the model to generate instructions on creating chemical weapons, and even planning terrorist attacks. Furthermore, DeepSeek has been accused of misrepresenting its training costs, with SemiAnalysis estimating that the company invested over $500 million in Nvidia GPUs alone, despite export controls. There are claims the US is investigating whether DeepSeek is acquiring these GPUs through gray market sales via Singapore.
ImgSrc: sm.pcmag.com
References:
- Mobinet AI: A Guide to Image and Textual Generative AI - Reports on DeepSeek's vulnerabilities and its ability to generate instructions on creating chemical weapons, and a terrorist attack. - 14d
- Pivot to AI - Details DeepSeek's issues: government bans, lack of guardrails, and cost misrepresentations. - 14d
- PCMag Middle East ai - The No DeepSeek on Government Devices Act comes after a study found direct links between the app and state-owned China Mobile. - 13d
- AI News - US lawmakers are pushing for a DeepSeek ban after security researchers found the app transferring user data to a banned state-owned company. - 13d
- mobinetai.com - Article on DeepSeek's ability to generate instructions for harmful activities, including chemical weapons and terrorist attacks. - 13d
- www.artificialintelligence-news.com - News article about DeepSeek's data transfer to a banned state-owned company and the security concerns that follow. - 13d