IMD AI Safety Clock makes biggest leap yet amid weaponization and rise of agentic AI
One year after launch, AI risk tracker underlines worrying implications of accelerating AI adoption....
Artificial Intelligence
Cybersecurity: Risks and opportunities in the scaling of industrial AI
Concerns over the ‘unpredictability’ of AI are widespread, but Industrial AI proves the value proposition of this tool, says Öykü Işık. This article explains how to build security into your strategy.
The use of artificial intelligence (AI) offers game-changing opportunities for businesses to improve operations and scale, and companies are rushing to embrace these chances – yet fears over potential security threats have tempered enthusiasm.
Scaling broadens the attack surface, making industrial systems more vulnerable to cyber threats. AI models can be targeted by adversarial attacks, data poisoning, or model inversion techniques that expose sensitive information. The increased connectivity of industrial AI systems creates additional entry points for hackers.
Yet, by proactively integrating security measures into AI development and deployment, organizations can minimize risks while maximizing the potential of this tool.
Relying on traditional machine learning (ML), rather than the less-well-tested generative AI, industrial AI has been used for tasks including predictive maintenance, quality checks, and energy management for more than 15 years.
The opportunity for industries to improve cyber resilience
AI is most successful when organizations have specific and well-defined use cases, they know what they are measuring, and they have proprietary data. Industrial AI, the application of artificial intelligence in business settings including manufacturing, energy, or construction, clearly showcases the potential of AI. Relying on traditional machine learning (ML), rather than the less-well-tested generative AI, industrial AI has been used for tasks including predictive maintenance, quality checks, and energy management for more than 15 years. ML relies on pattern recognition, learning from the past to predict current needs, and the efficiency gains are proven.
“AI introduces unique attack surfaces, requiring proactive and AI-specific security measures. Implementing zero-trust AI, robust data security, and adversarial defenses is critical.”
Industrial AI as a use case
As Cedrik Neike, managing board member and CEO of Digital Industries, Siemens, writes in a new report A New Pace of Change: Industrial AI x Sustainability: “In fact, industry has steadily been developing AI since the 1970s, making it reliable, secure, trustworthy, and suitable for industrial use. ‘Industrial AI’ now meets the requirements of the most demanding environments, enabling us to communicate with software, equipment, or machines in natural language, and helping us to design processes or even entire plants.”
Navigating the cyber threat landscape
Still, AI is changing the cyber threat landscape. AI introduces unique attack surfaces, requiring proactive and AI-specific security measures. Implementing zero-trust AI, robust data security, and adversarial defenses is critical. It is also crucial to remember that AI security is not just an IT issue, it’s a business priority affecting compliance, reputation, and trust.
Managing risk
Building reliable, accurate, and secure AI requires attention to all four elements of the traditional threat landscape, from actors to types to surface to core assets. AI introduces the prospect of unique attack surfaces within these four threat entry points, requiring proactive and AI-focused security measures.
Secure by design approach
To mitigate risks, when building industrial AI processes, you have to follow a secure-by-design approach; you cannot just fix issues as they come up. A secure-by-design approach integrates security into the design and building of products, systems, and applications, creating AI that is secure by default. A big aspect of this is the AI vendor: if the vendor builds ignoring vulnerabilities, you will bring those into your organization. For example, while there was widespread enthusiasm for the lower cost of DeepSeek, the fast and less expensive Chinese AI software challenger to Silicon Valley stalwarts like OpenAI, Google, or Meta, independent security evaluations of DeepSeek have found a series of weaknesses, and exposed susceptibility to cyber threats including prompt injection attacks, jailbreaking or data poisoning. Cheaper options may not represent savings in the long term.
AI security strategies
Mitigating cybersecurity risks when using AI requires a multi-layered approach that includes robust security measures, governance frameworks, and continuous monitoring. Here is a checklist for the five key strategies:
1. Secure AI model development and deployment
- Explainability
- Governance
- Developer training
- Secure software development
2. Adversarial AI and model manipulation
- Bias audits
- Adversarial training
- Input sanitization
- Validate training data
3. Data privacy and protection
- Encryption
- Privacy compliance
- EU AI Act
- Access control
4. AI supply chain and third-party security
- Vet external AI tools
- Security in the cloud
- Software bill of materials
5. AI-specific incident response and monitoring
- Continuous AI threat monitoring
- Incident response for AI failures
- Fair use policies
Authors
Related
AI: the key to human-centered business
Fears that AI will dehumanize work overlook its potential to make organizations more human. Here’s how organizations can harness the power of human-centered AI....
LLMs will hallucinate forever – here is what that means for your AI strategy
OpenAI’s research shows AI hallucination is unavoidable. Leaders must adapt strategies, using limits as opportunities for innovation and resilience....
Can AI ever be “just another teammate”?
As AI grows more capable, HR leaders must carefully weigh its impact on team dynamics before treating it as just another teammate, says IMD’s Ginka Toegel....