Why transparency is vital in the AI era
Mats Thulin, Director of AI and Video Analytics, Axis Communications, discusses the role the modern security-tech industry must play in ensuring the safe, effective deployment of AI.
AI technology is transforming the security landscape.
Advanced video analytics offer new opportunities to refine and enhance the security process.
AI is changing the rules: human-centric security is rapidly transforming into human-augmented security. Our growing knowledge of AI, combined with innovations in the use of multiple data streams and algorithmic analysis, means we’re entering a new era, one in which AI filters out irrelevant information, leaving security teams sharper and more effective than ever.
But let us be clear: for all its growing power, AI is not a universal solution. It is a vital aid and a great driver of efficiency, but it is not an answer to every problem. Without careful deployment AI can get things wrong. It is effective in the right situations, but blind reliance on AI can lead to costly mistakes, false alarms, or overlooked threats. Its performance depends on quality data, thoughtful configuration, and a full understanding of its limitations.
The burden and importance of clarity
Responsibility for this understanding goes beyond the user. As AI-driven security solutions become more prevalent, it is vital that manufacturers, vendors and installers highlight what they can and can’t do.
Transparency is essential for building trust and enabling end users to make informed decisions. The security industry benefits from a head start in understanding and managing the use of AI – and its potential misuse. With that experience comes a responsibility to ensure that AI is deployed ethically and effectively.
Technology providers, therefore, must generate trust through transparency
Responsibility does not just mean ensuring that technology is used properly, it means being open about the capabilities of AI, the areas in which it will shine and those in which it might struggle. Businesses can only make informed decisions on their security if they have the full information.
They must be able to grasp how different systems perform under various conditions, because any misrepresentation of capabilities or limitations can lead to dangerous blind spots.
Safety through solution design
Some use cases are more sensitive than others. For organisations in critical infrastructure, manufacturing, or chemical processing, an undetected security breach or a system failure could lead to severe consequences or loss of life. In any industry, an AI tool that misses the presence of a trespasser or a mechanical malfunction could result in significant harm.
No AI solution offers universal functionality – its abilities are influenced by the environment in which it operates. AI is, in every case, situational. Certain analytics, for instance, may struggle in poor lighting or noisy environments, necessitating additional data sources or different positioning. Poor camera sensors may not be able to offer AI engines the data they need to perform analytics well. And biases, too, can impact the decision-making capabilities of AI.
If a model consistently struggles to detect movement in certain weather conditions, for example, it is critical that users are informed. Clarity about the limitations of AI helps create a good foundation of overall system design and allows businesses to construct more effective security solutions by avoiding flawed assumptions about AI’s capabilities.
Ensuring responsible use of AI
Technology providers have a secondary responsibility. The power and versatility of AI systems mean they have, like any technology, risks of exploitation and misuse. Some may employ AI in unethical or illegal ways. The security industry has always carried ethical responsibilities in areas such as surveillance. With AI, those responsibilities are equally important.
Measures must extend beyond vendors refusing to sell to customers known to use technology irresponsibly
AI’s growth demands a transparent approach throughout the sales channel. It is imperative that vendors, integrators and manufacturers work hard to foster trusted, open relationships with each other – and exhibit the discretion to work only with partners that align with their ethical and moral views on AI use.
Building an AI future – the right way
The pace of AI innovation makes it difficult to predict what security technology will look like in the coming years. Given this uncertainty, the need for an ethical discussion cannot be overstated: by prioritising openness and fostering strong industry relationships, security providers can positively contribute to a safe and responsible future for AI. Manufacturers should place trust at the top of their list of business goals.
Trust creates a significant competitive advantage in a crowded market, as vendors and integrators are far more likely to recommend a company that has shown a commitment to responsible behaviour over one which acts impulsively over something as powerful and divisive as AI.
As artificial intelligence evolves, the businesses which have demonstrated such a commitment today will be the ones trusted in the future. They are the ones that will lead the industry forward towards a smarter, safer world.
Mats Thulin
Director of AI and Video Analytics
Learn more about Axis approach to AI: www.axis.com/campaigns/ai-in-video-surveiliance
