Is Customer Data Safe with AI Tools? A Guide for Nonprofits

The Nonprofit AI Gap: 92% Are Using It, 86% Have No Policy

AI adoption in the nonprofit sector has reached near-saturation, but the guardrails haven't arrived. A recent industry survey found that 92% of nonprofits now use AI in some capacity, while a separate sector statistics roundup reports that 86% of organizations have no AI policy governing how it's used.

That gap between use and governance is where donor data gets exposed. Seven in ten nonprofit professionals say they worry about data privacy and security when using generative AI, yet worry clearly isn't translating into written rules. Most AI work happens informally: 82% of nonprofits use chatbots on an ad hoc basis, and only 17% pay for a subscription. That means much of the sector's AI activity runs on free consumer accounts, which typically offer weaker data protections than managed business plans.

The stakes are higher for nonprofits than many realize. Security researchers describe the sector as "cyber-poor but target-rich" — organizations hold valuable donor records, financial data, and sometimes client healthcare information, while spending relatively little on cybersecurity. Up to 60% of nonprofits have already experienced a cyberattack or data breach in recent years, and the research links the sector's expanding digital footprint — cloud services, online platforms, and emerging technologies like generative AI — directly to increased exposure.

Several structural factors make this gap hard to close:

• 60% of nonprofits lack in-house expertise to assess AI tools, and only 4% have an AI-specific training budget.
• 81% of staff use AI individually on an ad hoc basis; only 4% of organizations have documented, repeatable AI workflows.
• Almost two-thirds of nonprofits report that none or just a few staff have a solid understanding of AI.
• 90% of foundation leaders provide no funding or support for grantees' AI use, leaving nonprofits to figure out governance alone.

The informal-use problem deserves particular attention. When a development associate pastes a donor list into a free chatbot to draft an appeal, that's not a strategy — it's a workaround, as one sector CEO put it. Each unmanaged interaction is a data-handling decision no one at the organization approved, reviewed, or documented. And because breaches in the sector often stem from human factors like phishing and misconfiguration rather than purely technical failures, unvetted staff behavior with AI tools fits the exact risk pattern.

The encouraging news buried in the data: privacy-conscious AI use doesn't sacrifice results. AI-powered donation forms that operate without personally identifiable information outperform industry averages, yielding $161 one-time gifts versus the $115 benchmark. Effectiveness and data protection aren't in tension — but closing the gap requires organizations to treat governance as the starting point, not an afterthought.

Where the Real Risks Live: Free Tools, Untrained Staff, and Vendor Blind Spots

When a staff member pastes donor records into a free chatbot account, the breach risk doesn't come from the AI itself — it comes from everything around it. Three exposure points stand out: unmanaged consumer tools, a sector-wide expertise gap, and vendor relationships nobody has vetted.

Start with the tools. 82% of nonprofits use AI chatbots informally, and only 17% pay for a subscription, according to sector research. That means most AI work happens on free consumer accounts, which typically offer weaker data protections and no administrative oversight than paid team or enterprise plans. When 81% of staff use AI individually on an ad hoc basis and just 4% have documented, repeatable workflows, there's no way to know — let alone control — where sensitive information is going.

Then there's the people problem. 60% of nonprofits lack in-house expertise to assess AI tools, and only 4% have any AI-specific training budget. Among nonprofit marketers actually using generative AI, 69% have received no formal training. This matters because the peer-reviewed cybersecurity literature describes nonprofits as "cyber-poor but target-rich" — and finds that breaches usually stem from human factors like phishing and misconfigured systems, not exotic technical failures. Untrained staff pasting beneficiary data into unvetted tools is precisely that pattern, amplified. It may help explain why up to 60% of nonprofits have experienced a cyberattack or data breach in recent years, according to peer-reviewed research on the sector.

Vendor blind spots compound the issue. With no expertise in-house, most organizations never read the data-handling terms, never ask who bears breach costs, and never require a breach-notification timeframe — the three contract provisions nonprofit privacy attorneys say matter most.

The trust dimension is just as exposed. Donors and funders are watching:

• 76% of donors want nonprofits to disclose when and how AI is used, but only 26% of nonprofits do so.
• 23% of foundations will not accept grant applications containing generative AI content, and 67% remain undecided.
• 90% of foundations offer no AI funding or support to grantees, leaving nonprofits to figure out safe adoption alone.

That gap between donor expectations and nonprofit disclosure is a reputational risk in its own right. And because nonprofits hold donor, financial, and sometimes healthcare data, a trust failure travels further than it would in most small businesses.

The encouraging part: none of this requires abandoning AI. Privacy-conscious designs already outperform — AI donation forms that never touch personally identifiable information average $161 one-time gifts versus a $115 industry benchmark. The fix is governance, not avoidance: managed accounts, a written policy, trained staff, and vetted vendors. For teams evaluating AI tools and agent skills, independent vetting resources help — SkillzBench, for example, is building a free directory that security-scans and benchmarks AI agent skills so organizations can see risk before adoption, not after.

The Five Safeguards That Actually Protect Donor Data

Here's the encouraging news buried in the sector's data-safety anxiety: the safeguards that actually protect donor data are simple, cheap, and mostly administrative. While up to 60% of nonprofits have experienced a cyberattack or data breach in recent years, breaches typically stem from human factors — misconfigured systems, phishing, social engineering — rather than exotic technical failures. That means governance, not technology, is your best defense.

Data privacy attorney Jon Dartley of Perlman & Perlman puts it plainly: nonprofits must adhere to data privacy laws and implement robust security measures to safeguard sensitive information while upholding donor trust. Based on converging guidance from nonprofit attorneys and sector leaders, here are the five safeguards that do the real work.

• A written AI use policy with no-PII-by-default rules. With 86% of nonprofits lacking any AI policy, this is the highest-impact step you can take. The core rule: no personally identifiable donor or beneficiary data goes into AI tools without proper legal agreements in place. Keep it short — most organizations need a few clear rules, not a formal strategy document.
• Human review of all AI output. Nothing AI-generated reaches a donor, grant application, or beneficiary without a person reading it first. This catches both privacy leaks and the accuracy errors that concern 63% of nonprofit professionals.
• Vendor contracts with explicit breach terms. Before signing with any AI vendor, identify your liability limit, specify in writing who bears responsibility and cost of a data breach, and require a specific breach-notification timeframe. These are the terms Dartley recommends negotiating in every contract.
• Consolidate staff onto managed accounts. 82% of nonprofits use AI chatbots informally, and only 17% pay for subscriptions — meaning most donor-adjacent work happens on free consumer tools with weaker data protections. Moving staff onto managed team plans with administrative control lets you actually see what data goes where.
• Proactive disclosure to donors and funders. 76% of donors say it's important for nonprofits to disclose when and how AI is used, yet only 26% of organizations do. And 23% of foundations won't accept grant applications containing generative AI content — so disclosure protects both trust and funding eligibility.

The most powerful proof point here is that privacy-conscious AI isn't a compromise. AI-powered donation forms that operate without touching personally identifiable information actually outperform industry averages — sector data shows an average one-time gift of $161 versus the $115 industry average, and $32 average monthly recurring gifts versus $24.

That finding reframes the whole question. Effective AI use doesn't require exposing donor PII, and as Nathan Chappell, chief AI officer at Virtuous, notes, the teams pulling ahead are the ones willing to establish simple guardrails and intentionally integrate AI into how decisions get made. The safeguards above are those guardrails — and none of them requires a budget line or a security engineer.

One practical note as you build these workflows: if you extend AI beyond chatbots into agents and automated skills, vet what you install. SkillzBench is building a free directory that security-scans and benchmarks agent skills so organizations can see which ones are safe before connecting them to their tools — a useful extension of the same no-surprises principle that governs donor data.

Your 30-Day Data-Safe AI Implementation Plan

Knowing the risks is one thing; building the habits is another. With 86% of nonprofits operating without an AI policy — while 70% of their staff worry about data privacy — a 30-day plan turns good intentions into a routine your team can actually follow.

Week 1: Draft a simple policy. Skip the formal strategy document. As one sector guide puts it, most nonprofits need just three to five repeatable prompts that solve real problems, not a 40-page playbook. Write down those workflows, and add the non-negotiables: no donor or beneficiary PII in AI tools, and human review of every AI-generated output before it goes out.

Week 2: Audit existing tools. You can't govern what you can't see. Survey data shows 82% of nonprofits use AI chatbots informally, and only 17% pay for subscriptions — meaning donor-adjacent work is likely happening on free consumer accounts with weaker data protections. Ask staff what they use and for what.

Week 3: Move to managed accounts. Consolidate informal usage onto team or enterprise plans with administrative oversight. This matters in a sector where up to 60% of nonprofits have experienced a cyberattack or data breach in recent years. Also flag upcoming vendor renewals so you can add breach terms — liability limits, who bears breach costs, and a required notification timeframe — as nonprofit attorney Jon Dartley recommends.

Week 4: Publish a donor-facing disclosure. 76% of donors want nonprofits to disclose how AI is used, but only 26% do. A short, honest statement on your website closes that trust gap — and protects funding, since 23% of foundations reject grant applications containing generative AI content.

Automating the PII-free workflows. Once your safe workflows are defined, agent skills can run them automatically — with one caveat: published research found roughly 26% of skills carry at least one vulnerability. SkillzBench is building free security scanning, version-pinned benchmarks, and a free MCP connector so nonprofits can adopt only verified skills, with a trust score and measured Lift shown separately for every listing. Browse the free directory at skillzbench.com and equip your agents with skills that are proven, not just popular.