Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

The risks of ‘cheating on everything’ with AI

While “cheat on everything” is certainly an intriguing and provocative value proposition, there are safer and more efficient ways to use and implement the latest AI tools.

The risks of “cheating on everything” with AI

[StockPhotoPro / Adobe Stock]

Profile photo of Dux Raymond Sy

Chief Transformation Officer · AvePoint

San Francisco-based startup Cluely made waves earlier this year when it raised $5.3 million in seed funding to build out its AI assistant that allows users to, as the company says, “cheat on everything” by providing hard-to-detect, real-time assistance to users on a wide array of tasks. For context, Cluely was founded by a Columbia University student who was suspended by the University for using the company’s product, which he built, to cheat on software engineering interviews (this is where the “cheat on everything” slogan comes from). The early success of Cluely’s product has already generated viral controversy online, with some claiming that the technology is dystopian, unethical, and even harmful to humans.  

At the end of the day, the Cluely saga is a timely reminder that AI should be used to amplify human engagement with the world, rather than “empowering” us to “cheat on everything.” Fundamentally, it’s a tool for good—but in order to realize this positive potential, companies need firm ethical and data governance frameworks, and they need to stick to them to guarantee long-term success with AI.

HOW COVERT AI USE CREATES GREATER RISKS

It’s no secret that AI presents a growing and unique set of data security and cybersecurity risks. AI-enabled threats like AI-driven malware, social engineering attacks, and prompt injection attacks have already spread around the globe, and the risk of accidental data exposure with AI is also immense. According to the 2024 AI and Information Management Report, for example, 46% of organizations experienced a data exposure event during AI adoption. 

The best way to mitigate AI-related data security and cybersecurity risks is to implement strong data governance policies across your organization. This includes strong information lifecycle management, stringent regulatory adherence, and the classification of enterprise data according to relevance and sensitivity (this ensures that irrelevant or confidential data isn’t accidentally shared by your organization’s AI assistant). Strong data governance for AI also includes establishing clear rules around AI use and misuse within your organization so that employees know how to act in a way that minimizes the risk of accidental data disclosures and limits exposure to malicious attacks.  

Acceptable use policies are an important part of every AI-readiness plan, and it’s crucial to understand that any deviation from these policies will substantially increase risks. Covertly using an outside AI assistant to engage in meetings, for example, would expose internal (and potentially confidential) business conversations to outside actors and organizations. While an intelligent AI assistant with real-time listening and output capabilities could be extremely useful, covert and unregulated/unsecured use of such a product in a business context presents serious risks. That’s why it’s important not to “cheat,” but instead to use AI in a regulated, secure way that limits vulnerabilities. In this case, cheating isn’t an ethical problem—it’s a legitimate cause for data security and cybersecurity concerns.

3 BEST PRACTICES TO DRIVE SECURE, EFFICIENT AI ENGAGEMENT

While it’s important to acknowledge the risks posed by the spread and use of AI technology, it’s also important not to regulate risk in a way that drives down engagement with AI tools. Some see the tension between productivity and security as an either/or choice (in other words, that you can only have productive AI use on the one hand, or secure use on the other) but in fact, it’s possible and widely advised to implement AI controls in a way that helps employees be more efficient.   

Organizations can follow these best practices to implement AI safely and securely:   

  • Implement Rigorous Data Segmentation: Divide organizational data into well-defined categories based on sensitivity, relevance, and confidentiality. Ensure AI systems access only non-sensitive and pre-approved data to safeguard critical business information from accidental exposure.  
  • Establish Comprehensive Training Programs: Educate employees on secure AI practices, emphasizing the risks of inputting proprietary or confidential data into external AI systems. Pair this with clear acceptable use guidelines to ensure employees understand and follow best practices while using AI tools.   
  • Enable Role-Based Access And Monitoring: Restrict AI usage to authorized personnel based on their roles and responsibilities, ensuring sensitive data is accessible only to those who need it. Combine this with regular activity monitoring to swiftly detect and address unauthorized behavior or anomalies. 

While “cheat on everything” is certainly an intriguing and provocative value proposition, there are safer and more efficient ways to use and implement the latest AI tools. By following these best practices and creating strong data governance policies across the board, you can make your teams more efficient without putting your organization’s data at risk.