AI Guide: Governance and Policies

Table of Contents

    Your newsroom stands to benefit from developing an AI policy, even if you’re not using AI. Why? Many people in your organization, including freelancers, may already be using generative AI in their work. And your audience may assume you’re using AI if you haven’t disclosed if or how you’re using the tech. Letting others fill in those gaps creates risk, but it’s risk that can be mitigated with clear, appropriate guidance that supports workers and maintains trust with audiences.

    Consider Your Approach

    Before you start evangelizing for AI, you should ask yourself: How comfortable are you with your staff or freelancers using AI in their work? What’s the worst-possible scenario? What’s the best case? Do you want to go as far as to require or preclude AI use?

    Consider these hypothetical situations
    • A new employee sends over a couple of story ideas that aren’t very compelling or interesting. When asked about it, the employee says they used AI as a thought partner, which they thought was allowed because they’ve seen others on the team use a similar application. 
    • A sales rep looks to adjust an advertising contract from one client to the next, so he uses an AI application. In the process, some of the client’s key data is uploaded to an application that’s training on input data. 
    • Your managing editor is using AI to do a first pass at the intern resumes she’s received for your summer program. The resulting pile is the least diverse set of candidates you’ve seen in years. 
    • A freelancer submits a draft story that includes odd-sounding quotes you suspect were invented by an LLM.

    What do you do?

    This is where AI governance and policies can play a strong role, but it's not always foolproof (or enforceable). You’ll need to make a decision about how much control you want to have over your team devising their own products or workflows.

    Start With Your Mission and People

    The best place to start with an AI policy is to look inward at your organization: its mission, values, priorities, and the audience you seek to serve. Any AI policy should make sense within this larger context and reflect your principles as a news organization.

    Step-by-step instructions (plus templates)

    Begin with what AI can do for your organization — rather than what it can’t.

    Will you use AI to deepen your reporting capabilities? Expand your reach to new populations? Give your audience more personalized content or better experiences? A newsroom that does a lot of investigative reporting may need a different AI policy, for example, than a community-based daily news organization. Consider your formats as well: An AI use policy for audio or television will look different from one focused on social media.

    Next, consider your audience and stakeholders.

    What does your audience think about AI? Will they understand the nuances of how AI can be used to strengthen or amplify journalism?

    How might your donors or subscribers respond to a published AI policy? What about the institutions that partner with or support your organization, financially or otherwise?

    You can take the temperature of these groups formally or informally. You may never get concrete answers to every question, but they are still worth considering as part of the larger picture.

    Think about your team and colleagues.

    To know where you’re going, you’ll want to know where you are. Are your employees, freelancers, and colleagues already using AI in their work? If so, how, and with what tools? This feedback can be collected through a simple form or a few small listening sessions. Newsrooms with more than 10 employees may want to use an anonymous survey that can be repeated annually – data that you can benchmark each year. You may also want to consider creating a task force or committee to help develop and review the policy.

    Here are some key things you’ll want to learn:

    • How do employees feel about AI? Do they view it with fear and trepidation, or with curiosity and a sense of opportunity?
    • What do they know about AI? Do they have a good handle on how it works? Do they understand its strongest and weakest use cases for their jobs?
    • Are they already using AI in their work, and if so, how? When they have questions about acceptable use, what do they do? Who do they ask?

    If you find yourself stuck in this process, try flipping the script again. Ask your team what they believe AI cannot do in journalism. What will always remain sacred about human-produced journalism?

    Templates to help you get started:

    Research and Draft It Out

    The next step is to research how similar organizations have approached their AI policies. Pay particular attention to how each policy is framed, which uses of AI are considered acceptable, and where the organization draws clear boundaries.

    Examples and databases you can use

    Here are several examples and databases to get you started:

    As you review the many policies available, you’ll notice that a strong newsroom AI policy typically includes:

    • Specific examples of when AI may be used in journalism
    • Specific examples of when AI may not be used
    • Clear standards for disclosing the use of AI
    • A practical definition of what it means to keep a “human in the loop”
    • A reaffirmation of the organization’s commitment to human-led reporting and writing, accountability, accuracy, and audience trust

    Consider the Risks

    This technology is not without its risks. A simple search will yield many examples of news organizations that have regretted instances of AI use. And while it’s impossible to know every minefield in this technology, publishers can start by considering four areas of risk.

    Four areas of risk
    • Reader trust: How will our use of AI impact the trust that readers have in our work? What are their expectations for our work? What kind of error would damage their trust irreversibly? What kind of risk are we willing to take on in our designated use case?
    • Privacy and Confidentiality: What kinds of safeguards do we need to feel comfortable about putting our work – from reporting to subscriber data – in an LLM? What happens if something goes wrong? What’s our liability?
    • Bias: Many studies show LLMs carry biases against certain cultures, races, and genders. How could this impact our AI use cases? What should we do to minimize this?
    • Costs: How much will this new technology cost your organization? How much are you willing to pay in tokens, credits, or service charges to ensure security and proper guardrails? Will you put limits on how much AI is available to readers and your team?

    Now, consider each of these questions for each area of risk:

    • What’s the worst possible scenario? Think of the worst possible situation you can imagine due to AI – intentional or unintentional error. 
    • How likely is that scenario? Put a number on it, if you can, knowing your staff, resources, and work. 
    • What guardrails could be put to mitigate this risk? What safeguards or policies could you put in place to decrease risk?

    Get comfortable with the idea that, because AI is a nondeterministic and evolving technology, you’ll never have certainty. Your goal is to take that probability as reasonably close to zero as you can while balancing the need to push your organization forward.

    Debate the Details

    There is no shortage of nuance in these policies, so it is important to talk through the details with your staff. For example, if your policy says that AI cannot be used for writing of any kind, does that also prohibit using it to generate headline ideas or draft alt text for photos? If your policy says you only use human-produced images, what about the editing (through AI or another technology) to adjust illustrations for social media?

    More questions to ask

    Questions like the above are why you should circulate the policy among your team and invite feedback, ideally in writing. Ask:

    • Are you comfortable with this policy?
    • Does it reflect our mission and values as a news organization?
    • How might readers respond to it?
    • How could it affect their trust in our journalism?

    It can also be helpful to gather feedback from other stakeholders, even informally. An audience group, advisory board, or community partner may identify concerns or ambiguities that your internal team has overlooked. That feedback can help you produce a clearer, more coherent, and more public-ready policy.

    Once you have a draft, consider several practical questions:

    • How will the policy be enforced?
    • How will it be communicated internally and externally?
    • Who will be responsible for updating it?
    • What developments or circumstances should trigger a review?

    Finally, include the date the policy was last updated. Your approach will change as the technology, your organization, and audience expectations evolve — likely sooner than you think.

    What about a technology policy?

    An AI policy should be distinct from your organization’s existing technology policy (if you have one). The two may overlap, but they serve different purposes.

    What each policy should do

    What each policy does (or should do):

    Technology Policy

    • Explains how employees may use workplace tools and technology, including which tools are approved by the organization

    • Covers data security, privacy and appropriate use of specific tools

    • Applies broadly to the organization’s technology stack

    • Is typically intended for internal use

     

    AI Policy

    • Defines permissible and prohibited uses of artificial intelligence

    • Covers AI wherever it appears, including within existing systems such as Google search summaries or productivity software

    • Applies specifically to AI, which the policy should clearly define

    • May also be published for readers and other external stakeholders

     

    Next Steps

    • Leave room to grow. AI is developing quickly, so plan to review and update the policy every few months. When drafting it, consider not only what AI can do today, but what its capabilities may look like in the near future. No policy will be completely future-proof, but understanding what is emerging can help you avoid writing rules that are outdated almost immediately.
    • Let the policy find the people – not the other way around. An effective policy cannot live only in an email, shared drive, or PDF on your website. It should remain visible and accessible to employees. Consider posting it in the office, reviewing it regularly during team meetings, and creating ongoing opportunities for staff to ask questions or raise concerns. Ideally, that cadence should also include continued training, even if some of it is self-guided.

    Revisit Your Reader's Perspective Through the Trust Lens

    Before you hit publish (or send) on your new AI policy, you’ll want to revisit again the paramount concept of reader trust.

    More on prioritizing trust

    Trusting News, a research and training non-profit for journalists, surveyed readers on how they view AI disclosures in news – from the position of the disclaimer, to how much detail was included, and where. The results showed that disclosure and transparency were the most important factors, but that presentation (length and disclosure of AI usage) also mattered.  

    Put yourself in the shoes of a few members of your audience: 

    • The most skeptical reader: What will they think when they read this policy, see an AI disclosure on a story, or note you’ve used AI to help personalize their news experience? 
    • Your most loyal readers or supporters: Will they feel that your use of AI cheapens your value, or augments it? 
    • Your casual or future reader – someone you want to bring into your publication’s community. How will they see your use of AI if this is the first time they visit your website? Will it matter to them?

    Their (imagined) feedback should not dominate your policy, but you must still anticipate their questions. Keep in mind that they might not care – or even support you – if it means you can deliver more of the journalism that they’ve come to rely upon.

    Additional Reading and Resources

    Return to Guide Home

    Generative AI was used to refine the writing and check content on this page. All content was created, reviewed, and verified by a human, who takes full responsibility for its accuracy.