Companies can develop AI solutions in niche or specific areas like healthcare, finance, manufacturing at a faster rate and lower cost. The DeepSeek Mixture of Experts (MoE) approach allows companies to mix and match experts, basically picking the model experts they need and contributing to the development of an expert module to solve a problem.
DeepSeek is gaining traction globally for the unparalleled control it offers businesses to build new products and internal solutions. But the freedom to build comes with its own risks, leaving businesses with the challenge of managing ethical concerns around data privacy and security.
According to Danny Hammo, Founder and CEO of ExecAI, the opportunities of DeepSeek hinge on developers choosing the right architecture and data management systems.
Big opportunities: "Companies can now run AI projects at scale without breaking the bank," Hammo says. "Instead of selecting a few projects due to budget constraints, companies can now afford to run tens of projects, experiment with new ideas, and let their developers loose using DeepSeek."
Adding to the reasons companies should embrace DeepSeek he adds, "Companies/organizations, especially the ones sensitive to data privacy, can run a fork of DeepSeek in their own data centers and machines. This gives them absolute control, which allows them to build in-house products and solutions to tackle some of the internal problems, mundane tasks, automation, data analysis, etc."
"Companies can develop AI solutions in niche or specific areas like healthcare, finance, and manufacturing at a faster rate and lower cost. The DeepSeek Mixture of Experts (MoE) approach allows companies to mix and match experts, basically picking the model experts they need and contributing to the development of an expert module to solve a problem."
The risks: While many remain hopeful about DeepSeek’s open-source model, Hammo sees it as a potential risk. "Like any open-source project, the continued success of DeepSeek depends on the community of contributors and developers. If the project is not supported by enough researchers and developers, model releases will lag, issues might take long to solve, etc."
Instead of selecting a few projects due to budget constraints, companies can now afford to run tens of projects, experiment with new ideas, and let their developers loose using DeepSeek.
Options for data security: Security is top of mind for most developers and companies who want to work with DeepSeek. In light of all the potential opportunities for companies, Hammo adds that developers have options to tackle security risks.
- Self-managed: "Developers today have a few options with DeepSeek," Hammo explains. "They can clone DeepSeek source code from GitHub and run on self-managed compute (local/home computers or servers). The developer has to ensure that their servers are secure, they have to properly randomize and obfuscate private data, and they have to properly store the data to avoid leaks. This is the responsibility of the developer and requires a lot of work on the developer end."
- Cloud-hosted: "Hosted version of DeepSeek on cloud providers, like AWS, is another approach." Hammo suggests. "With this, data security is managed by the cloud provider, and they provide the tooling for data privacy. In my opinion, this is the best approach."
- Direct API: "There is direct DeepSeek API use, most likely a managed instance of DeepSeek on a Chinese compute provider. The developer is relying on the provider’s data privacy and security mechanisms. Generally speaking, privacy and data security should be a priority if developers intend for their products to succeed. The balance, in my opinion, lies in the developer choice even if they are planning to use OpenAI or Google Gemini—the developer is bound by their terms and service."
