Organizations looking to create value from generative AI are increasingly relying on their data.Properly harnessed,these corporate assets can unlock efficiencies and new revenue streams,crucial for gaining competitive advantages.Consulting firm McKinsey&Company predicts that generative AI could boost annual profits by$2.6 to$4.4 trillion,prompting many companies to explore AI applications for crafting sales and marketing materials,brainstorming digital products,and writing software code.

　　However,as with any emerging technology,generative AI services come with their share of risks and uncertainties.Running AI solutions in your data center can offer more flexibility and control over data.Here are three keys to curbing generative AI’s security risks:

　　1.Weigh Generative AI’s Many Risks

　　Control is crucial as concerns about managing generative AI technologies grow.Many employees already use text,image,and video generators to enhance their workstreams,often without IT’s approval,leading to the rise of shadow AI.

　　Generative AI’s output can inadvertently disclose information from its training data.Large Language Models(LLMs)can inherit biases or generate inaccurate content,posing ethical and legal risks to your business.Additionally,LLMs can be exploited to create malicious code,such as phishing scams and malware.Generative AI services are vulnerable to prompt injection attacks,where attackers extract prompts and any files employees have uploaded to models.Moreover,bad actors have become adept at creating realistic deep fakes.

　　Data used to train LLMs may require additional protection to prevent unauthorized disclosures.Data sovereignty rules govern how data is stored,processed,and used globally.These challenges make it difficult for IT leaders to weigh the risks of adopting generative AI.A recent Dell survey revealed that 45%of IT decision-makers cited data and IP risks as the primary reason for hesitating to embrace generative AI.

　　2.Build and Run On-Premises to Reduce Risks

　　While there’s no silver bullet solution to protecting corporate assets,training and augmenting a generative AI model on-premises can help reduce bias,disinformation,and other threats to data and corporate reputation.Think of your on-premises data center as a bank vault protecting sensitive IP,allowing you to prioritize security,privacy,governance,and control.

　　Employing a zero-trust strategy can further enhance data protection and defend corporate assets.This strategy allows you to control who has access to data,what data is shared,and how to remediate security issues.Zero-trust principles also enable you to encrypt and monitor data while instituting governance and communication policies for responsible use.

　　3.Generative AI in a Multi-Cloud World

　　While on-premises solutions may offer the best security and customization options for generative AI models,some organizations may require a broader,more portable approach.This might involve running generative AI apps in multiple operating environments,close to where the data resides.Organizations may build applications on-premises but run them in private or public clouds for agility or at the edge where lower latency is crucial.

　　A recent Dell survey showed that 82%of IT decision-makers are interested in an on-premises or hybrid approach to building their generative AI solution.Identifying the best place to securely run generative AI workloads can be challenging,and many organizations may lack the talent to do so.This is where an ecosystem of partners becomes valuable.

　　Dell Technologies offers a comprehensive portfolio of generative AI solutions,including client devices and server infrastructure that support AI workload requirements.By applying a zero-trust approach to architecture,Dell is well-positioned to help secure your applications and data as you adopt generative AI.

　　Learn more about how Dell views the intersection of generative AI and cybersecurity.

　　This post was created by Dell Technologies with Insider Studios.