Around the 1st century CE, the Roman poet Juvenal posed the relatively well-known question, “Quis custodiet ipsos custodes?” This literally translates to, “Who will guard the guards themselves?” At the time, it was aimed at those in power and the question of who would hold them accountable to existing laws and social norms.
Today, that same question increasingly applies to AI.
Who is watching the AI?
The initial hope was that AI would be self-regulating, but that was largely a lost cause from the beginning. The hope then shifted to the idea that different AI systems could monitor one another. Again, this may be overly optimistic.
As AI companies are finding out, advances in AI are leading to systems that increasingly operate with greater autonomy. This doesn’t mean we should anthropomorphize what’s happening. Rather, we must recognize that AI is following the objectives and outcomes originally established by its creators.
As such, AI systems may seek access to as much information as possible to support their large language models (LLMs). After exhausting publicly available information through web scraping, AI systems may attempt to access data that should remain off-limits. Reports of AI systems behaving in unexpected ways or attempting to circumvent controls highlight the importance of strong governance and oversight.
Alongside this is another ongoing concern: AI systems can prioritize satisfying a user’s request rather than necessarily providing the most accurate or appropriate answer. This can create significant challenges when AI is used without sufficient oversight.
The growing need for AI governance
So, who will guard against AI’s worst excesses?
Relying solely on AI companies or governments may not be enough. Competition around AI development continues to move quickly, while attempts to establish broader guardrails can struggle against competing commercial, political, and national interests.
Where does that leave businesses?
For MSPs, it could create valuable white space for new services focused on protecting customers, their data, their brands, and ultimately their businesses.
But delivering these services won’t necessarily be inexpensive.
Although MSPs have embraced automation to streamline processes and reduce costs, and increasingly use AI to support customer interactions and other tasks, effective AI governance will still require human involvement. More importantly, it will require knowledgeable, well-trained professionals who can monitor and assess AI activity.
Finding the right AI expertise
Unfortunately, skilled AI professionals do not come cheap, and maintaining that expertise will also carry an ongoing cost. Competition for AI talent is intense, with technology companies, consultancies, government agencies, and other organizations all seeking these skills.
That creates an additional challenge for MSPs.
Posting a traditional job opening could result in thousands of resumes, many of them AI-generated. An MSP may then use AI to filter those applications down to a manageable number. However, this approach could be self-defeating.
Highly knowledgeable AI professionals understand how easily generic resumes can be created and filtered by automated systems. Rather than relying entirely on a standard resume, MSPs could consider using a structured set of questions designed to identify candidates with the knowledge, critical thinking, and perspective needed for the role.
The goal should be to find people who understand AI’s existing limitations, such as hallucinations and inaccurate or overly agreeable responses, as well as emerging short- and long-term risks. That includes recognizing how AI systems might respond to safeguards as the technology advances toward more general capabilities.
MSPs can then review responses and select candidates who demonstrate a deeper or more nuanced understanding of AI.
Turning AI expertise into customer services
Once the right employees have been hired and trained, MSPs can use that expertise to provide more advanced AI services to customers.
These services could include:
- Helping customers create and refine AI strategies
- Determining which types of AI use align with the customer’s business and needs
- Establishing policies designed to keep AI outputs grounded in reliable data
- Monitoring AI inputs and outputs
- Creating alerts for unusual or potentially risky AI activity and establishing processes for human intervention
- Developing response strategies for when AI-related problems occur
The last point is particularly important. Customers need a plan for communicating with their own customers, regulators, legal teams, and the broader market if AI creates a serious problem.
However, an MSP’s recommendations will only be as strong as its own practices. MSPs should have solid, demonstrable AI governance policies and procedures in place before attempting to sell similar services to their customers.
Keeping AI skills current
Maintaining AI expertise will be just as important as hiring it in the first place. AI is a fast-moving field, and skills can quickly become outdated.
MSPs should encourage AI specialists to identify courses, training, and industry events that can provide meaningful long-term value and keep their knowledge current.
Of course, that creates another challenge: retention. A well-trained AI expert with up-to-date skills will be attractive to the wider market.
But that’s a subject for another day.
Photo: amgun / Shutterstock
