Let’s talk about DeepSeek.

It’s one of the latest AI tools generating buzz, that got many talking, and even more trying it out.

I mean, why wouldn’t they? It’s extremely powerful, and it’s free!

It allows you to do what other paid AI tools can, without subscriptions or commitments.

Plus, it comes ready with the R1, an AI model that can provide quality responses with deeper, structured thinking.

On the surface, it’s a game-changer.

But dig deeper, and the risks pile up fast.

Over the last few weeks, some of my clients and contacts asked if DeepSeek is too good to be true, and if it’s safe for corporations and businesses.

Well, as the saying goes, there’s no free lunch—especially when it comes to AI tools. Free access often means paying with your data, security, or compliance risks.

And we’re about to find out the true cost of using this free AI tool.

Still, this article is based on my research into this AI tool, and is solely based on uncovered facts and my personal experiences. As a user of the tool, I have nothing against DeepSeek, and I believe it has enabled more people to access “intelligence” for an incredibly low cost.

Of course, as an AI enabler, L&D consultant and Generative AI trainer, I must be mindful of the harm it will bring to my corporate and SME clients.

By the end of this piece, you will have a better understanding of the risks and better awareness of what’s really at stake when using free AI tools.

Also, the version of DeepSeek I’m referencing throughout this article is the version you can access through your web browser and their app. I’m excluding those who are interfacing with DeepSeek’s AI model through API or those who are self-hosted.

DeepSeek on Data Privacy

DeepSeek on Data Privacy: Who Sees Your Data?

Here’s the big one: DeepSeek’s web version stores all user data—including chat histories, file uploads, and personal details—on servers in China. That means anything you type in, whether it’s a strategic business question or a sensitive HR document, could be transmitted abroad.

And under Chinese law, the government can compel companies to hand over data upon request.

Translation? If you’re using DeepSeek for corporate matters, you might be exposing confidential information to state-level scrutiny and even potential industry espionage.

And it doesn’t stop there:

• Broad Data Collection: DeepSeek gathers not just your inputs, but also profile information (email, phone number), IP address, and even keystroke patterns. It may also share this data with affiliates or advertisers.

• Lack of Transparency: While users can delete chat histories, it’s unclear how long DeepSeek retains data. There’s no clear policy on anonymization or data minimization.

• Ownership Ambiguity: Once your data is in DeepSeek, who owns it—you, your company, or DeepSeek? The answer is murky, and that’s a legal risk waiting to happen.

Bottom line? If your business deals with proprietary, regulated, or sensitive information, using DeepSeek is a gamble.

DeepSeek on Security Issues

DeepSeek on Security Vulnerabilities: A Cyberattack Waiting to Happen?

Security researchers have uncovered serious flaws in DeepSeek’s security protocols. Let’s run through the worst offenders:

• Weak Encryption: Investigations found outdated encryption methods, including the broken Triple DES (3DES) cipher. Even worse, DeepSeek’s iOS app disabled Apple’s App Transport Security (ATS), meaning some data was transmitted unencrypted over the internet. That means if an employee accesses DeepSeek over public Wi-Fi, hackers could intercept their data.

• Data Leaks & Breaches: DeepSeek has already suffered a major security lapse, where a publicly accessible database leaked over a million log entries, including user chat histories and API keys. If confidential business conversations or employee data were exposed in such a breach, the consequences could be severe.

• Suspicious Data Flows: Some analyses found DeepSeek transmitting data to domains linked to Chinese state-owned companies and embedding undisclosed code from ByteDance (the company behind TikTok). There are also anti-debugging mechanisms in place—an unusual move that makes it harder for security professionals to inspect how the app really works.

Your data isn’t just vulnerable to leaks. It could be actively exploited. Some cybersecurity experts have even recommended deleting DeepSeek from enterprise and government devices altogether.

DeepSeek on Compliance Risks

Compliance Risks: Will DeepSeek Get You in Legal Trouble?

If your company operates in a heavily regulated industry (finance, healthcare, legal, etc.), DeepSeek could be a compliance nightmare due to strict data protection laws, industry-specific security requirements, and potential legal liabilities tied to data sovereignty and privacy regulations.

• GDPR Violations: The European Union has strict rules about exporting user data outside the EU, especially to China. DeepSeek openly stores all user data in China but provides little transparency on compliance safeguards. Regulators in Italy and Ireland are already investigating its legality.

• HIPAA Non-Compliance: For healthcare organizations, DeepSeek is a non-starter. It does not have the necessary certifications to handle protected health information and lacks essential agreements like HIPAA Business Associate Agreements.

• Government Bans & Restrictions: Countries like Australia, Taiwan, and South Korea have already restricted DeepSeek on official networks. In the U.S., a proposed bill seeks to impose multi-million-dollar fines on companies using Chinese AI in critical sectors.

If you’re in a regulated industry, using DeepSeek could expose you to fines, legal action, or a PR disaster.

DeepSeek on AI Bias & Ethical Concerns

DeepSeek on AI Bias & Ethical Concerns: Can You Trust DeepSeek’s Responses?

AI tools need strong safeguards to ensure ethical, unbiased, and responsible use. DeepSeek struggles in this area.

• Biased & Dangerous Outputs: DeepSeek has been found to be 11 times more likely to generate harmful content compared to Western AI models. Researchers have successfully manipulated it into producing detailed instructions for money laundering and malware creation—things most AI models would refuse to do.

• Censorship & Political Bias: DeepSeek follows China’s strict censorship rules. Ask it about politically sensitive topics, and it will either deflect or provide government-approved narratives. If your business values open discussions and factual neutrality, this is a dealbreaker.

• DEI Risks: AI bias can impact workplace diversity, equity, and inclusion. If DeepSeek has unchecked biases, it could generate gender-biased hiring advice, culturally insensitive responses, or incomplete viewpoints—a potential liability for HR and training teams.

DeepSeek on Reliability & Performance

DeepSeek on Reliability & Performance: Can You Count on It?

Even if we ignore all the security, privacy, and compliance risks (which you shouldn’t), DeepSeek still struggles with stability and reliability:

• Frequent Downtime: The web version of DeepSeek has experienced multiple outages, including a major one in February 2025. If you’re using it for work and it goes down, you’re out of luck.

• Hallucinations & Inaccurate Information: Like all AI, DeepSeek sometimes fabricates facts. The difference? It does so more often than its competitors. Businesses relying on its output without verifying accuracy could make costly mistakes.

• Immature Platform: DeepSeek is still evolving, and its support and troubleshooting infrastructure are weak. Unlike established AI providers, it lacks a public bug bounty program or strong enterprise support.

Why we don’t use DeepSeek as the main tool during our workshops?

Is It Safe To Use DeepSeek During Training Workshops

When conducting AI-related training programs, whether it’s on prompt engineering or Generative AI in general, some external trainers and training providers might use DeepSeek for in-class activities. It might be for cost-cutting measures or a demonstration of the capabilities of a cheaper AI model, but the risks are ever-present.

Which is why in our Workplace AI Mastery programs, while we do showcase DeepSeek’s amazing ability, we do it after letting the participants know the risk of this powerful AI model.

That’s because we believe the best way to navigate uncertainties, is by the insights and awareness that comes through education. Only by learning where the dangers are, can we truly steer away from it.

And for the rest for the prompting practices, whether it’s during or after the workshop, they will be using our own AI tools, powered by safer models like Antropic’s Claude or OpenAI’s GPT-4o.

So, what should we do with DeepSeek?

Final Verdict: Should You Use DeepSeek?

DeepSeek is undeniably powerful, but it comes with massive trade-offs—significant privacy risks, security vulnerabilities, compliance concerns, and reliability issues that could turn a convenient tool into a liability.

If your work involves proprietary, regulated, or sensitive data, you should think twice before using it.

Here’s what I would recommend:

✅ Use it for experimentation, NOT for critical business decisions.

✅ NEVER input confidential, financial, or personal data.

✅ Monitor security developments & regulatory scrutiny.

✅ Consider alternatives with stronger compliance and security.

The AI revolution is exciting, but due diligence is non-negotiable. The last thing you want is for a tool meant to increase efficiency to turn into a liability. Proceed with caution!