AI breakthrough or security threat? Singapore firms assess DeepSeek’s impact as data privacy concerns persist

SINGAPORE, Feb 9 — Companies in Singapore are cautious about using DeepSeek despite its cost-saving potential, citing concerns over data security and AI biases.

The Straits Times reported that major firms, including banks, consulting agencies and cyber-security companies, have set clear rules prohibiting employees from using generative AI tools like DeepSeek for work without due diligence.

“The model stands out for being open-source, performing well across benchmarks and offering a significantly lower-cost alternative to competitors,” said Hanno Stegmann, managing director and partner of Boston Consulting Group’s (BCG) AI team.

DeepSeek gained attention in the tech sector when it launched its R1 model in January, claiming it could rival OpenAI’s technology at a fraction of the cost.

The Chinese AI start-up said the R1 model cost S$7.6 million (RM24.8 million) to train, significantly less than the hundreds of millions spent by US tech giants on their large language models (LLMs).

Following its release, R1 topped download charts and caused US tech stocks to drop, reflecting concerns about competition for established generative AI players like OpenAI, Google and Amazon Web Services.

Singapore-based AI consumer insights platform Ai Palette estimated that using DeepSeek could cut infrastructure costs by 40 to 60 per cent, particularly in high-end computing chips required for large-scale LLMs.

BCG noted a surge in client interest in DeepSeek but advised caution before deployment.

“It is worth waiting for a more thorough assessment of DeepSeek’s risks before using the model,” said Stegmann.

Concerns about generative AI tools like OpenAI, Google and Amazon Web Services have been raised since their launch two years ago, with key issues including data retention and risks of corporate data leaks.

Early tests show DeepSeek does not meet some responsible AI standards.

Tech enthusiasts flagged that the model avoids answering questions on topics often censored by the Chinese government but provides responses on similar issues in other countries.

Experts also raised concerns about the DeepSeek app retaining prompts and results to train the model.

It remains unclear how much data is stored and scrutinised.

“We strongly recommend our clients assess any potential risks before using DeepSeek, especially regarding data security and confidentiality,” said Stegmann.

Evaluating DeepSeek involves repeated tests to assess answer consistency, potential misuse and biases, such as gender or demographic-related issues.

“It is fair to say that first releases of many LLMs had some issues at the beginning that had to be ironed out based on user feedback and changes made to the model,” he added.

South Korea and several European countries have requested information on DeepSeek’s training data.

“It’s worth waiting to see what additional information may emerge on the data DeepSeek was trained on,” said Stegmann.

South Korea, Italy and Australia have blocked access to DeepSeek on government devices due to security concerns.

These restrictions resemble those imposed on ChatGPT in its early days in 2022, when governments temporarily blocked access before reassessing its use. ChatGPT remains blocked in China.

Legal firms in Singapore are also taking a cautious approach.

RPC tech lawyer Nicholas Lauw said generative AI tools are prohibited for handling client data until their safety is thoroughly tested.

“Our stance is precautionary, designed to maintain the trust and integrity of our client relationships, and aligns with wider regulatory guidance and best practice,” Lauw said.

RPC is testing LLMs to determine which can be securely and effectively deployed for internal use.

These AI models undergo legal risk assessments and checks on accuracy, ethics compliance and sensitive data exposure risks.

Several major firms have partnered with AI developers like Microsoft and OpenAI to create customised AI tools that ensure corporate data remains secure.

OCBC Bank and UOB use internal AI chatbots for coding and archive searches. OCBC staff laptops restrict access to external chatbots, including DeepSeek, unless they meet security requirements.

“Our generative AI applications are built using on-premise, open-source LLMs to minimise data leakage risks. We evaluate all new LLMs prior to making decisions on deploying them,” said Donald MacDonald, head of OCBC’s group data office.

Rajesh Sreenivasan, head of tech law at Rajah and Tann, said the firm has invested heavily in enterprise editions of Microsoft Copilot and legal assistant Harvey AI, ensuring data remains within its systems.

“This is especially important for regulated firms to ensure sensitive data is not ingested by AI developers for training,” said Sreenivasan.

Enterprise deals also include indemnity clauses that protect corporate users from legal risks, shifting responsibility to tech vendors.

Most major generative AI vendors, including Microsoft, IBM, Adobe, Google and OpenAI, provide indemnity protections for clients.

“DeepSeek doesn’t have an enterprise product yet,” said Sreenivasan. “It might be open-source, but this alone doesn’t protect corporate users from potential legal risks.”

Despite these concerns, some firms have started using DeepSeek.

Babbobox chief executive Alex Chan allows employees to use various AI models, including DeepSeek, for productivity tasks like finding inspiration or coding.

Tony Zhu, chief technology officer at conversational AI solutions provider Wiz.AI, sees potential in using R1 for text-based customer support.

Wiz.AI has been using DeepSeek’s technology since December 2024, citing its capability in complex reasoning tasks. The firm also utilises open-source models like Meta’s Llama.

Somsubhra GanChoudhuri, co-founder of Ai Palette, said DeepSeek’s advancements could help smaller firms in Singapore adopt AI despite budget constraints.

DeepSeek’s cost-saving techniques could also inspire local firms to experiment and innovate with generative AI, even without large-scale computing resources, said Kenddrick Chan, who leads the digital international relations project at think-tank LSE Ideas.

The Ministry of Digital Development and Information told The Straits Times on February 7: “The Government does not generally comment on commercial products. We advise companies to evaluate products on their own merits and the risks of use, including compliance with relevant laws.”

Sreenivasan said DeepSeek’s entry will drive competition in the market. “We’ll see even more of a contest in the generative AI space, and that’s good for innovation.”