How To Create An Effective Ai Usage Policy

This is an excerpt from Chapter 2 of “AI for nan Enterprise: The Playbook for Developing and Scaling Your AI Strategy,” a caller ebook by acclaimed tech journalist Jennifer Riggins and sponsored by Red Hat and Intel.

From nan advantages of utilizing nan “two-speed” AI finance model, to measuring nan existent effect of AI, this free book, now disposable for download, helps endeavor leaders create an AI strategy to unlock productivity gains, lick antecedently intolerable problems and summation a existent competitory edge.

Most organizations needed an AI usage argumentation yesterday. Yet only astir half person them.

“How galore products person had AI shoved successful location while wholly ignoring nan superior symptom points of their products and their platforms?” asked Hannah Foxwell, laminitis of AI for nan Rest of Us. “Your AI strategy needs to beryllium grounded successful nan business today, not successful immoderate fictitious type of nan business tomorrow.”

To create your AI policy, found a cross-functional, cross-organizational AI enablement squad — possibly seated successful nan caller CAIO office.

The International Organization for Standardization’s AI standards are a bully spot to commencement considering your AI policy:

• ISO/IEC 42001 – AI governance and consequence management: Establishes accountability and oversight, mounting up consequence guidance protocols to identify, entree and mitigate imaginable risks and impacts.
• ISO 27001 – Information security: Recommends tooling that vets crossed people, processes and technology. This includes a privateness guidance extension.
• ISO 37301 – Compliance and ethical culture: Defines requirements for compliance guidance systems, arsenic good arsenic guidelines to nurture a civilization of integrity, honesty and fairness successful AI usage.

The European Union’s AI Act offers different bully measurement to categorize consequence astir AI tooling:

• Unacceptable AI: Includes scraping nan net aliases information cameras to build facial nickname databases.
• High-risk AI: Includes immoderate AI for captious infrastructure and utilities, acquisition and employment management.
• General-purpose AI models: For AI systems that are transparent by default, which allows users to make informed decisions.

This categorization whitethorn alteration for different parts of your business, depending connected their levels of consequence acceptance aliases avoidance.

It’s besides wise to springiness everyone successful your statement a refresher successful section and world information privateness regulations, for illustration nan EU’s GDPR and nan California Privacy Rights Act (CPRA). Generative AI (GenAI) tools’ easiness of usage and take person already led to inadvertent information leakage — you don’t want to beryllium next.

Communicate Your AI Policy

Once your AI enablement squad agrees connected an AI usage policy, make judge you pass it pinch each stakeholders wrong and extracurricular nan organization.

Grammarly has been utilizing AI successful its merchandise for nan past 15 years, and its User Trust Center has go an manufacture modular for clear AI communication. It breaks nan risks down into 4 areas:

• Privacy
• Security
• Compliance
• Responsible AI

“Managers and leaders are going done this learning travel alongside their teams, truthful you can’t look to your head for actual advice,” Foxwell said. “A centralized enablement squad aliases outer training partner is simply a bully measurement to spell if you want to get accordant messaging.”

Don’t hide that your AI strategy must besides facet nan SaaS products you are already using. If they’ve injected immoderate benignant of AI into nan product, has your institution vetted it for compliance?

Since these risks and usage cases alteration by domain, you can’t person 1 argumentation to norm them all. Consider immoderate overarching data and information rules, but past let taxable matter and domain experts to measurement successful connected department-specific aspects. And crushed immoderate policies successful examples based by domain aliases department.

As you create this AI usage policy, conscionable retrieve to “translate” it from legalese to laic position truthful that each labor understand. This is simply a bully usage lawsuit for GenAI — truthful agelong arsenic you person a quality reviewer successful nan loop.

To publication more, download “AI for nan Enterprise: The Playbook for Developing and Scaling Your AI Strategy” today!

Group Created pinch Sketch.