Common staff concerns about AI

This guide gives leaders and staff shared language, data, and narrative options for conversations they are probably already having about AI. Use the language that fits your organization, and leave what does not.

The one-page version

• Concerns about AI often come from a genuine place. Take them seriously rather than dismissing them.
• Opting out does not pause AI. It will keep shaping the world either way.
• The better question is: where can careful use help animals enough to justify the cost and risk?
• Leaders should say the organization’s position out loud: what the org has decided, why, what the boundaries are, and how people can raise concerns.
• Used well, AI is also an opportunity for people — a chance to grow capacity as an advocate and move toward higher-value work, not only a risk to manage.
• Humans stay responsible. AI can help with the work, but people still make the calls and answer for the claims.

Each row has a full section below: what people say, what is underneath it, what the data shows, and language options.

Why this exists

If you lead a team of animal advocates, AI has probably become one more thing you’re expected to have answers about. A lot of people have mixed feelings about it. We do too.

These tools can help small teams draft grants, summarize research, translate materials, analyze data, and plan campaigns. They also raise real concerns about energy, jobs, privacy, copyright, labor, bias, accuracy, and trust.

We should take those concerns seriously. We should also be honest about the other side of the choice. If animal advocates opt out by default, AI will still shape the world. Companies, governments, political campaigns, and animal agriculture will keep using it. The question for us is whether we can use these tools carefully enough to help animals, protect trust, and stay aligned with our values.

How to use this guide

• Proactively, before concerns surface. This is the use we recommend most. Unaddressed concerns do not go away; they go quiet and harden. Naming the concerns early, before anyone has to raise them, changes the whole conversation.
• When someone raises a concern in a meeting or chat. Read that concern’s section before responding, and start with the “what might be going on” layer. These conversations go better when people feel heard before they hear data.
• When a board member, funder, or partner asks where you stand. The one-page version and “The narrative choice” give you a position you can state in two minutes.
• When rolling out a new tool or policy. Walk through the concerns most relevant to the affected team and address them in the announcement itself, not after the pushback.

Start with the real value

A lot of AI conversations begin with concerns. That is understandable, but it can make the whole topic feel defensive. There is a positive case too: these tools can help animal advocates do more useful work with the same small teams.

• A researcher can turn a stack of reports into a first-pass brief, then check the sources and sharpen the conclusion.
• A fundraiser can organize messy notes into a grant draft, then rewrite it with human judgment and funder context.
• A communications team can test several versions of a message before asking which one is clearest, most accurate, or most likely to backfire.
• An operations team can reduce repetitive admin work: meeting summaries, intake triage, status updates, and follow-up drafts.

These results are already showing up in the movement. A research organization completes some social science projects about 25% faster, with human review of methods and final claims. A food systems organization automated roughly 90% of a repeated data-analysis workflow, letting one analyst support far more clients.

One example from our own work: the first run of Amplify for Animals, our 12-week AI training program, was built and delivered by four part-time people. They used AI to brainstorm and plan the program, pitch it, stand up a new website and intake process, and create new slide decks and brand-new content for hundreds of advocates every week, for 12 weeks in a row. The program would not have been possible without it — and there are similar examples from every role across the movement.

None of these wins came from a flashy demo. They came from repeated help on real work. Strong AI use usually looks boring from the outside: a person with good judgment uses a tool to get unstuck, see more options, or move faster on work that already mattered.

For the version of this organized by job, see the role guides — what AI means for campaigners, fundraisers, researchers, operations staff, and more.

The narrative choice

People will create a story about AI whether leaders provide one or not. If leaders do not name a story, people often fill in the blanks themselves: layoffs, surveillance, pressure to do more with less, or implied permission to use risky tools. Naming a clear story will not erase the concerns, but it gives people something honest to respond to.

The story we tend to use is responsible opt-in. AI is not automatically good, morally clean, or safe. But blanket refusal is not neutral either — it means giving up real potential impact for animals. It can leave animal advocates with less capacity while animal agriculture, corporations, campaigns, and governments keep using these tools. Factory farming is one of the largest systems of suffering and environmental damage in the world. Careful AI use helps us challenge that system more effectively.

A simple version of the stance: use AI where it helps animals; protect people and sensitive data; keep human judgment central; avoid deception; watch the evidence; and change course when the facts change. We use six pillars to put that stance into practice: leadership, policy, training, tools, workflows, and ongoing learning.

If you lead an organization, make the stance explicit. Say what the organization has decided for now, why it matters for the mission, which uses are encouraged, which uses are off-limits, and where concerns should go. People do not need to feel the same way about AI, but they should know how the organization is approaching it. Vague silence creates more anxiety than a clear position people can respond to.

A useful personal framing is the opt-in versus opt-out calculation: “Everyone has to choose their own relationship to AI. The way I think about it is: what happens if I opt in, and what happens if I opt out? If I opt in, I get access to a tool that can help me make more of a difference for animals. If I opt out, AI continues anyway, other actors keep using it, and I give up some capacity to help.” Adapt the language to fit your role and your organization; do not repeat it if you do not believe it.

When the conversation starts to sprawl, one line can help bring it back to the mission: “A lot of the concerns about AI are real. The question for us as advocates is: what do we do about it?”

Use the data carefully

AI data changes quickly. Estimates on energy, water, jobs, privacy, copyright, and reliability are still moving, and researchers disagree on some of them. Use numbers to ground discussion, not to pretend the debate is settled. Before publishing exact claims, check the linked sources and the date.

In the spirit of the guide: the data below was compiled with AI assistance and then spot-checked by a person against the original sources. That is the pattern we recommend everywhere — the tool does the gathering, a human verifies before anything is published. The sources here were last verified in June 2026.

Concern 1: Environmental impact

What someone might say: “AI is terrible for the environment. Every prompt uses energy and water. How can we justify using this while claiming to care about animals and the planet?”

What might be going on: This is often a real climate concern mixed with a fear of hypocrisy. People see posts about data centers, water, and energy, and the moral conclusion can feel obvious. There can also be social pressure — in some circles, using AI has become a quick signal that someone is careless or ethically compromised.

What seems true: AI has a footprint. Data centers use energy and water, and local impacts can be serious. That should not be hand-waved away. At the same time, ordinary text use by animal advocates is not the same question as global AI infrastructure buildout. For scale: a typical chatbot prompt uses roughly 0.3 watt-hours of electricity — about one second of microwave use — and somewhere between a few drops and a few tablespoons of water, depending on what the estimate counts. An hour-long video call uses a few hundred prompts’ worth of electricity, and even at the most expansive water estimates, a single shower uses more water than 500 prompts. The status quo has a footprint too: animal agriculture is a major driver of emissions, land use, water use, and biodiversity loss. Whether AI has a cost is settled — it does. The live question is not whether each prompt is worth it — at the individual level the numbers are tiny — but whether using these tools well helps shrink a much larger system of harm.

Useful data:

• Food production accounts for over a quarter of global greenhouse gas emissions, and more than three-quarters of agricultural land is used for livestock, even though meat and dairy provide only 18% of calories. Our World in Data
• The IEA projects global data center electricity use will more than double to around 945 TWh by 2030 — roughly Japan’s total electricity consumption today, and a little under 3% of global electricity use. Local impacts can be much larger. IEA
• Google measured a median Gemini text prompt at 0.24 Wh of electricity and 0.26 mL of water — about five drops. Third-party estimates that count electricity generation and older models run higher, into the tens of milliliters. Google
• Hannah Ritchie’s interactive comparison tool helps put AI prompts in context against everyday electricity uses, including a median ChatGPT query estimate of about 0.3 Wh. Hannah Ritchie
• Analyst Andy Masley argues that individual chatbot use should be separated from data center infrastructure questions, estimating a typical prompt at about 0.3 Wh — comparable to running a microwave for one second — and about 2 mL of water once electricity generation is included. Andy Masley

Language that may help:

• “I agree the environmental concern is real. I don’t think the answer is pretending AI has no footprint. I think the answer is using it only where the value is high enough.”
• “It’s true that AI uses energy and water — so does most of the technology we rely on. The question is whether the benefit to animals outweighs the cost, and we think it does.”
• “AI has an environmental impact, but it is small next to the footprint of animal agriculture. If careful use helps us challenge that system, the benefit to animals far outweighs the cost.”
• “Let’s separate three questions: wasteful use, useful day-to-day use, and the global buildout of data centers. They don’t all have the same answer.”

Practical next steps: Keep the conversation at the right level. In aggregate, AI will have a real environmental impact — mostly on the energy side, less so on water — and it is fair to care about that. An individual advocate’s use is a rounding error next to it, and next to the impact of the work itself. We do not recommend asking staff to weigh the footprint of each prompt: that deliberation costs more than it saves, and the mission impact of the work AI helps you do outweighs its footprint many times over. Save the scrutiny for organizational choices and public claims, and update environmental claims as the estimates change.

Concern 2: Job loss and replacement

What someone might say: “Leadership says AI will help us, but it feels like the real goal is to replace people or squeeze more work out of fewer staff.”

What might be going on: This is not paranoia. People have seen companies use new technology as cover for layoffs or workload creep. Staff may also worry about junior roles disappearing or career paths getting weaker.

What seems true: Work will change: some tasks will shrink, some roles will look very different in a few years, and some organizations will handle the transition badly. Saying “AI will not affect jobs” will not feel honest. But replacement is not the only path. Animal advocacy organizations usually have more useful work than capacity. Used well, AI can reduce repetitive work, help staff learn faster, and let teams take on projects that would otherwise sit untouched.

The more useful conversation is about role evolution. When a tool absorbs some of the repeatable work in a role, the question becomes what higher-value human work that person moves toward — reviewer, workflow designer, trainer, cross-functional integrator. The skills that make someone good at a role today are usually the same skills the AI-forward version of that role needs; what is often missing is that the organization has not yet named the next version of the work. This also shapes hiring: less weight on producing first drafts by hand, more on judgment, verification, and the ability to direct and check a tool. The role guides walk through this seat by seat.

Useful data:

• BCG estimates that 50% to 55% of US jobs may be reshaped by AI in the next two to three years, while 10% to 15% could be eliminated further out. BCG notes this is not an unemployment forecast. BCG
• A 2023 International Labour Organization study found that most jobs are only partly exposed to generative AI and are more likely to be complemented than substituted. ILO

Language that may help:

• “We are not adopting AI to replace you. We’re adopting it so the same team can do far more for animals.”
• “It would be dishonest to say work won’t change. It will. Our job is to shape that change around mission, people, and judgment.”
• “We don’t have a shortage of important work. We have a shortage of capacity. AI should help us do more of the work that matters.”

Practical next steps: Do not lead with “efficiency” if people will hear “layoffs.” Talk about capacity, learning, and animal impact. If your organization can honestly commit to not cutting staff because of AI, say so plainly and early — it is the single most reassuring thing a leader can say. If you cannot make that promise, do not make it; be honest about role evolution instead. Name role evolution explicitly and early — treat it as a management responsibility, not an individual side project. Protect junior staff development by teaching people how to think with AI, not only how to delegate to it.

Concern 3: Data privacy and security

What someone might say: “I don’t want donor data, legal strategy, staff information, or campaign plans going into some AI system. This feels risky and vague.”

What might be going on: This concern often means the organization has not given people clear rules. When staff do not know which tools are approved or what happens to the data, caution turns into blanket rejection.

What seems true: This is one of the strongest concerns because the risk is real and preventable. Drafting from public information is different from uploading donor records, HR notes, or campaign plans. “Never use AI” fails, and so does “use whatever tool you want.” What works is a simple policy people can follow: a tier of data that is safe to use freely, a tier that needs an approved tool or review, and a tier that stays out of AI tools entirely. The full treatment — the data tiers and the questions to ask a vendor — is in Data security and privacy basics, and the specific “does my data train the model” question has its own FAQ.

Useful data:

• Major providers say they do not train models on business data by default for their business and enterprise tiers, though this is provider-specific and settings-dependent, not a universal rule. OpenAI
• Simon Willison’s “lethal trifecta” is a useful warning: AI agents become especially risky when they combine private data, untrusted content, and the ability to communicate externally. Simon Willison

Language that may help:

• “You’re right to be cautious. We shouldn’t put sensitive data into random tools.”
• “Data security isn’t an argument against all AI. It’s an argument for approved tools, clear categories, and training.”
• “Safe use should be easy. Risky use should be clearly off-limits.”

Practical next steps: Adopt a simple data-tier policy and name which tools are approved for which tier (start from Data security and privacy basics). Be especially careful with AI agents that combine private data, untrusted content, and the ability to act externally. For most teams, human approval before external actions is the right starting point; remove it only deliberately, with guardrails you trust.

Concern 4: Big Tech, labor, copyright, and values

What someone might say: “AI feels exploitative. It’s built by huge tech companies, trained on people’s work, and dependent on hidden labor. Using it feels like betraying our values.”

What might be going on: This is often a values concern before it is a policy concern. Staff may have friends who are artists or writers, or be in activist spaces where AI is treated as theft. “This feels gross” is not a complete argument, but it is a signal worth listening to.

What seems true: There are unresolved ethical and legal questions around training data, worker treatment, corporate power, and creative labor. Individual abstention is a valid personal choice, but it does not solve the structural problem. An organization can acknowledge the discomfort, choose tools carefully, avoid harmful uses, and still decide that careful AI use is justified by the potential to help animals.

Useful data:

• The US Copyright Office has an ongoing initiative on AI-generated works and the use of copyrighted material in training; these questions remain unsettled and context-dependent. US Copyright Office

Language that may help:

• “I understand why this feels uncomfortable. We shouldn’t ask people to pretend the discomfort is irrational.”
• “Our values should shape how we use AI. They don’t automatically require us to avoid it in every case.”
• “We can be critical of the industry and still use these tools carefully for animals.”

Practical next steps: Do not present AI-generated creative work as purely human-created. Decide deliberately, as an organization, where you want to use AI in the creative process and where you do not, rather than letting it happen by default. Use human review for public-facing work. When there is a choice, prefer tools with stronger privacy, safety, labor, and governance practices.

Concern 5: Bias, hallucinations, quality, and trust

What someone might say: “AI makes things up, repeats bias, and produces bland work. I don’t trust it.”

What might be going on: A person might be reacting from professional pride, fear of public mistakes, or bad past experiences. It gets worse when people see AI drafts shared without review, as if confidence were the same thing as accuracy.

What seems true: AI can make up facts, flatten voice, miss context, and reproduce bias. It can also help draft, compare, summarize, brainstorm, and find gaps. The difference is whether people treat it as an assistant whose work gets reviewed or as an authority whose word gets trusted.

Useful data:

• Stanford HAI’s 2026 AI Index reports that hallucination risk depends heavily on the task: on a knowledge-recall benchmark designed to probe what models do not know, hallucination rates across 26 top models ranged from 22% to 94%; on a summarization benchmark, top models hallucinated in roughly 2% to 5% of cases. Risk varies enormously by task, which is why review and source-checking matter. Stanford AI Index 2026

Language that may help:

• “AI gets things wrong. It’s useful when a person stays responsible for checking it.”
• “The answer to hallucination risk is source-checking and review, not pretending the tool is useless.”
• “We wouldn’t publish an intern’s first draft without review. We shouldn’t publish an AI draft without review either.”

Practical next steps: Check sources for factual claims. Use AI for options and first drafts, not final authority. Use extra review for public, legal, financial, HR, donor, or strategic content. Give AI examples of the organization’s real voice before using it for communications.

Concern 6: Human judgment and overreliance

What someone might say: “I worry AI will make us lazy, less thoughtful, less creative, or less human. I don’t want our movement to outsource its judgment.”

What might be going on: A person might be trying to protect the craft of advocacy. People do not want advocacy to become shallow or automated, and may fear losing expertise, creativity, or moral seriousness.

What seems true: Overreliance is real. AI can weaken thinking if people use it to avoid thinking. It can also strengthen thinking if people use it to generate alternatives, test assumptions, critique drafts, or learn faster. Which way it goes comes down to habits, and habits can be trained.

Useful data: No single statistic settles this concern — it is primarily a training, habit, and leadership question.

Language that may help:

• “The goal isn’t to think less. It’s to think with better tools and keep responsibility with people.”
• “AI should help us ask better questions, not make final decisions for us.”
• “A good use of AI can make our thinking sharper: it can critique a draft, generate counterarguments, or expose blind spots.”

Practical next steps: For high-stakes work, ask what was checked and how. Use AI to create alternatives and critiques, not only finished outputs. Teach staff when not to use AI: moral judgment, the work of building relationships, and final responsibility stay human.

Concern 7: Hype and shallow use cases

What someone might say: “I’ve seen the AI demos. They look flashy, but I don’t see how this actually helps my work.”

What might be going on: This can be healthy skepticism. People are tired of new tools, bad trainings, and vague promises, and may be asked to learn AI without time, support, or a use case that touches their actual work.

What seems true: Generic AI adoption often fails: a demo does not change habits, and a tool subscription does not create value by itself. Value usually comes from picking real workflows, giving people time to practice, and checking whether the work gets better.

Useful data: BCG’s research suggests only about 5% of organizations have achieved substantial financial gains from AI so far — a useful reminder that value is not automatic, and comes from how people change the work, not from technology deployment alone. BCG

Language that may help:

• “You don’t need to be impressed by demos. Let’s test this on real work and see whether it helps.”
• “We’re not doing this because it’s trendy. The point is to find a few places where it clearly increases capacity for animals.”
• “Skepticism is useful if it helps us choose better workflows and measure real value.”

Practical next steps: Pick one or two real workflows before asking a team to adopt a tool broadly. Create regular practice time, not just one-off training. Share wins and failures. Measure something that matters — quality, time saved, reach, or staff load — and stop what does not help.

Concern 8: Reputation, transparency, and public trust

What someone might say: “If people find out we use AI, they may think our work is fake, lazy, or less trustworthy.”

What might be going on: A person might be trying to protect credibility or understand when disclosure is expected. It can also reflect a good instinct: trust is hard to build and easy to lose.

What seems true: Transparency matters, but the bar shifts with context. An AI-assisted internal outline is different from synthetic media, donor communications, or public evidence. The rule underneath all of it: do not deceive people, and keep humans accountable.

Language that may help:

• “We should be transparent where AI use changes what a reasonable person would expect about authorship, evidence, or authenticity.”
• “The standard isn’t disclosing every grammar suggestion. It’s no deception and clear accountability.”
• “We can use AI behind the scenes while keeping public claims, strategy, and voice in human hands.”

Practical next steps: Disclose AI use when it materially affects authenticity, evidence, or audience expectations. Do not use AI to fabricate sources, testimonials, or images of real events. Keep humans responsible for public voice and final claims. Create a short disclosure policy so staff do not have to guess — the AI policy guide includes a disclosure section.

Language that usually backfires

• Avoid: “AI has no environmental impact.” Better: “AI has a footprint. Let’s use it where the value is worth the cost.”
• Avoid: “AI won’t affect jobs.” Better: “Work will change. We want to shape that change around mission, learning, and people.”
• Avoid: “Everyone should use AI for everything.” Better: “Use approved tools for the right workflows, with review and judgment.”
• Avoid: “If you’re worried, you’re anti-progress.” Better: “The concern makes sense. Here is how we are thinking about the tradeoff.”
• Avoid: “This is mainly about saving money.” Better: “This is about increasing capacity for animals and reducing unsustainable work.”
• Avoid: “The data is settled.” Better: “The evidence is changing. Here is what we know right now.”

What now: a simple operating model

The next step after narrative is practice. We organize that work into six pillars — narrative and leadership mandate, policy and guardrails, training and adoption, tool selection and security, workflows and roles, and an ongoing rhythm that keeps adoption moving after the kickoff energy fades. The AI policy guide turns the guardrails pillar into a document you can adopt.

Where this leaves us

Concerns about these tools are normal, and often reasonable. You do not need everyone to feel enthusiastic for careful adoption to work — you need clear eyes, good rules, real use cases, and a habit of checking what helps animals.

For animal advocates, the central question is practical and moral: does this help us reduce suffering and increase impact while staying aligned with our values? If yes, we should learn to use it well. If no, we should not use it just because it is new.

Leaders across the movement are working through these same questions right now. If your team wants help, this is exactly what the Vegan Hacktivists AI Services program is for, at no cost.