The data breach at Chinese AI start-up DeepSeek has raised serious concerns about security in the rapidly evolving AI industry.
With sensitive user data, API keys and system logs reportedly left exposed, the incident has triggered warnings from global regulators, including Taiwan and Italy.
As AI adoption accelerates, the breach highlights the growing need for robust cybersecurity measures to protect critical data.
Industry experts share their insights on the implications of the DeepSeek breach, the challenges of securing AI platforms, and what businesses must do to safeguard their systems against future threats.
Dr Ben Goertzel, Founder of SingularityNET, said: “Decentralised databases don’t automatically solve all of AI’s problems, but they do push us closer to an AI ecosystem that is more secure, user-controlled and censorship-resistant, which is an important step forward.
“The shift from centralised AI data monopolies to decentralised architectures won’t immediately create more intelligent AI, but it will create a more open and accountable AI infrastructure, which is just as crucial.
“The advantages of decentralised AI databases – such as enhanced data ownership, privacy preservation, and fewer single points of failure – are clear. However, they don’t inherently address core AI challenges like hallucinations, limited compositional reasoning, or self-awareness.
“AI models running on decentralised infrastructure still face the same cognitive limitations as their centralised counterparts. Overcoming these challenges requires fundamentally different AI architectures than those underlying systems like OpenAI or DeepSeek.
“Around the world, R&D teams are actively exploring such architectures, including SingularityNET’s Hyperon project. One potential outcome of DeepSeek’s achievements could be that it prompts the world to take seriously the possibility of progress toward AGI – even without the need for billions of dollars in compute power.”
Neil Roseman, CEO of Invicti, said: “DeepSeek’s emergence in the AI landscape represents a predictable evolution rather than a watershed moment. Like previous computing advances, AI capabilities are becoming smaller, faster and cheaper at a pace exceeding expectations. DeepSeek’s superior price-to-performance ratio serves as a reality check for the AI industry, particularly U.S. companies and their venture capital backers, highlighting that return on invested capital matters in an industry demanding enormous resources without delivering revolutionary applications.
“We’re still in early stages of understanding AI’s value, with gains plateauing and improvements falling short of expectations. This mirrors Robert Solow’s observation about the PC revolution: ‘You can see computers everywhere but in the productivity statistics’. Claims about imminent artificial general intelligence remain far from reality, and tech companies citing AI to justify staff reductions are mostly addressing their excessive hiring during boom times, when large engineering teams were status symbols regardless of actual value.
“DeepSeek’s success reminds us that excessive spending without proportional returns is unsustainable. While companies make massive bets on AI, current results don’t justify these investments. Success will come from efficient, focused development addressing genuine needs.”
Sheldon Monteiro, CPO, Publicis Sapient, said: “The excitement around DeepSeek is understandable, but it’s important to separate immediate market reactions from long-term impact. The AI industry has a habit of overestimating short-term breakthroughs while underestimating their cumulative effects over time.
“DeepSeek’s efficiency gains are significant, but they follow a familiar pattern of optimisation in technology. This isn’t a revolutionary moment – though it may feel like a step change – it’s another refinement in the ongoing push for lower costs and higher performance, much like past advances in compute efficiency, model compression, fine-tuning techniques and parameter-efficient architectures.
“More notably, DeepSeek’s open-source approach will accelerate efficiency gains across the board, but the bigger shift is what it signals: proprietary advantage in frontier models is eroding. We’re moving toward a future where AI leadership is less about sheer model power and more about differentiated applications, domain-specific fine-tuned models and ecosystem integration.
“For businesses, the challenge isn’t keeping up with every new model release – it’s embedding AI in ways that create a real, defensible advantage. Yes, companies should stay flexible, ensuring they can switch models to capitalise on improvements in cost and performance. But the real winners won’t be those chasing the latest efficiency gains; they’ll be the ones integrating AI seamlessly into workflows, products and customer experiences to drive meaningful impact.”
Matt Calkins, CEO and Co-Founder, Appian, said: “The DeepSeek announcement is a wake-up call, and I would expect a bunch more surprises to come. Everybody thought they knew where the value came from. The war for AI is going to be an asymmetric war between the US, China and anybody else who gets involved. The things that make the US strong are not the same that make China strong. Therefore, the US should pursue its own path rather than trying to match China’s approach.
“DeepSeek is a great AI model that uses far less compute and expertise than we expected. Perhaps it has access to more compute than we realise, or perhaps it shows there was less ‘magic’ in compute and expertise than we expected.
“In my opinion, we’re going to see a commoditisation of AI. Many companies will achieve competitive AI, and a lack of differentiation will be bad for high-spending first-movers.”
Mike Britton, CIO, Abnormal Security, said: “DeepSeek’s claims of remarkably low costs are causing a stir in the industry, but the market’s reaction is based on taking the company at its word. Without a detailed view into costs and metrics, there remains uncertainty about whether their approach is truly faster or cheaper, or how the model was even trained. While it’s making waves now, a major backlash could follow as the more regulated parts of the world may be hesitant, or outright unwilling, to engage with it.
“Right now, much of the concern around DeepSeek is how it might threaten the current AI market with a competitive, cheaper alternative. But what’s also concerning, especially for the general public, is its potential for misuse. Bad actors are already using popular generative AI tools to automate their attacks. If they can gain access to even faster and cheaper AI tools, it could enable them to carry out sophisticated attacks at an unprecedented scale.”
Dan Shiebler, Head of Machine Learning, Abnormal Security, said: Smaller open source LLMs have been close on the heels of larger closed source LLMs since Meta’s first release of llama. The pace of closed source model performance has seriously plateaued since GPT4, and it’s not surprising to see smaller open source models catching up.
“It is certainly exciting that this model may have been trained with a tiny fraction of the compute required to train comparable models. I’m looking forward to seeing other labs replicate this training methodology.
“One thing to keep in mind is that benchmarks can be misleading. Although llama3 was announced as beating GPT4 when it was first released, very few users would describe llama3 as near-GPT4 parity at the context window lengths required for enterprise application. It remains to be seen whether Deepseek R1 is able to displace OpenAI O1 in real world workflows.”
