Understanding the threats of AI as you build your business

Artificial intelligence (AI) presents numerous opportunities for companies to streamline operations, uncover insights and enhance customer service.

There has been considerable fanfare surrounding AI, specifically generative AI chatbots, such as ChatGPT and Bard, that have become readily accessible to end users and are growing in sophistication.

In fact, it's undeniable that AI has become a prominent part of regular, everyday discourse amongst most people these days.

However, amidst the growing popularity and evolving technology that underlies AI, business leaders must take a clear-eyed view of the potential risks involved with widespread and unsupervised use. At present, with such an alarming lack of regulation and control around its influence and adaptability, there are real-world threats and risks that must be mitigated. Implementing AI securely requires proactive risk management from the outset.

This short article outlines what business leaders must do to ensure that their use of AI is ethical, moral, and economically beneficial to their companies. It will outline the most overt risks associated with AI use, and how leaders can mitigate them when sensibly integrating this technology into their firms. This has definitely grown in necessity if businesses do not want to be left behind.

The promises and perils of AI adoption

AI-enabled technologies like machine learning, natural language processing and computer vision are designed to enhance efficiency, productivity and decision-making.

According to PwC research, AI could contribute up to £232 billion to the UK economy by 2030. There is clearly a market for innovation which is hard to ignore. However, AI systems can also amplify risks if not thoughtfully implemented.

Certain high-profile examples of inherently biased algorithms and data breaches risk undermining public trust in organisations that are caught in the crossfire. Without addressing these risks upfront, the affected companies won’t see the true benefits of AI and automation and instead will be viewed as another statistic.

Data security risks

The underlying data utilised by AI systems is largely open-source, implying it’s readily available on the internet. So, as such, it’s safe to assume that some people digesting that information will not be using it for ethical reasons.

There have already been reports of unwarranted customer data exposure through AI, and thus, it’s fair to say that security breaches will continue to happen. This means that more stringent data protection controls must be adopted.

Here are some summarising points that business leaders must take note of:

• AI training data contains hugely valuable business insights that could be invaluable to competitors. Ensure data governance policies are in place and training data is anonymised

• Attackers are developing ways to steal or corrupt data used by AI through data poisoning and backdoor attacks. Keep a vigilant eye on data flows

• Open-source generative AI chatbots can be exploited to exfiltrate data. Adopt robust cyber security to safeguard your infrastructure

Breaches can expose sensitive customer and business data. If you handle sensitive data, you should exercise more stringent controls, particularly if your business sits in highly regulated industries like healthcare and finance.

Algorithmic bias and fairness

If algorithmic bias goes unchecked, AI risks automating and amplifying unfair outcomes. Businesses must take stringent steps to ensure that their adopted AI systems can be used without any fear of bias and contribute to a fairer and more equitable organisational culture.

Historical biases in data can lead algorithms to make discriminatory decisions about credit, jobs, healthcare and even personal characteristics. Audit data sets for evidence of bias and ensure rigorous testing before deployment, to ensure all personnel are protected.

Opaque AI models can produce biased results without clear explanations, which in turn generate misleading outputs that unsuspecting readers might perceive as factual. Businesses must ensure careful use of AI for unregulated and unsupervised outputs, to ensure that misinformation is not unwittingly perpetuated.

Cyber security vulnerabilities

AI expands the attack surface available for cybercriminals to capitalise on. It is widely believed that a large-scale network of criminal organisations aims to exploit AI for malicious purposes. It is therefore imperative that businesses adopt AI sensibly to avoid becoming unsuspecting victims.

• Attackers are developing AI-powered hacking tools, like using ‘deepfakes’ for social engineering. This level of sophisticated human impersonation poses serious threats to organisations practising everyday cyber security

• Vulnerabilities in AI code, infrastructure and development pipelines open new attack vectors and new opportunities to execute hacks on business infrastructure. Prioritise secure coding practices and go beyond the minimum requirements for data preservation and cyber resilience

• AI chatbots and customer support agents like ChatGPT must be hardened against potential misuse. Exercise supervision over escalating conversations and vet your logs for anomalies or suspicious activity

Risks of rogue or uncontrollable AI

AI programmes are purpose-built to use machine learning to refine and improve their efficiency, accuracy and validity. While the idea of superintelligent AI is largely theoretical and conjecture at this stage, any business that adopts AI – however marginally – has a responsibility to ensure it is integrated thoughtfully and not left to manifest without supervision.

Some top-level risks include:

• The capability of AI systems to act autonomously opens the possibility of unintended behaviour. Businesses should maintain human oversight and control measures

• Truly autonomous machines may make harmful independent decisions. Keep humans accountable for AI system outcomes

Best practices for secure, ethical AI

Responsibly harnessing AI's potential requires proactive risk management. Business leaders should:

• Perform extensive risk assessments before deploying AI and continuously monitor for emerging threats

• Implement comprehensive data governance policies, cyber security controls and access management

• Form diverse teams trained in AI and its use to oversee company-wide policies and practices

• Audit any programmes, algorithms and software regularly for bias

• Maintain human oversight and decision-making accountability over AI systems

• Embed principles like transparency, fairness and human benefit into AI model development

• Foster a culture of responsible AI through training and education across the organisation

• Collaborate with regulators and policymakers to shape emerging standards around ethical AI

AI holds tremendous promise for businesses looking to scale and grow but also poses a collection of legitimate risks if not carefully considered.

By taking a measured approach focused on security, transparency and ethics, business leaders can tap into AI's benefits while building consumer and stakeholder trust.

Relevant resources

• How to use generative AI responsibly: Key considerations for small businesses

• Harness the power of AI – business.connected Digital series

• Automation made easy