Update: AI Bot Attacks Did Access Some Donor Data

(Image Via Pexels.com)

Details regarding the artificial intelligence (AI) bot attack on multiple nonprofit fundraising platforms last week are starting to emerge, although many service providers are declining to comment or are not returning messages.

Two vendors were leaders are providing transparency are DonorPerfect, where the bots entered through marketing pages, and GoFundMe, which detached from Salesforce after the bots successfully entered the GoFundMe Pro area.

“Our investigation found that an unauthorized third party has made repeated attempts to access GoFundMe Pro and was able to gain access to personal donor and supporter information,” a GoFundMe spokesperson told The NonProfit Times. “We’ve taken immediate steps to secure the platform and conduct an investigation along with external cybersecurity professionals. At this time, we have addressed the known methods of unauthorized access and we are actively monitoring for suspicious activity. We have also notified law enforcement.”

GoFundMe originally reported that it disconnected from Salesforce as a precaution. According to a later statement from Salesforce to The NonProfit Times, “Salesforce disabled the GoFundMe Pro connected app and associated managed packages as a precaution after identifying suspicious use of access tokens associated with the third-party provider. This was a security measure taken to protect customers, not the result of a Salesforce platform service disruption.”

The statement from Salesforce continued: “The issue did not arise from a vulnerability within the core Salesforce platform. Salesforce and GoFundMe are working closely together to investigate this activity and safely restore the affected integrations as soon as possible. The integrations will remain unavailable until necessary security remediation is complete, and GoFundMe has demonstrated that the integrations can be safely restored.”

GoFundMe is notifying involved individuals in two ways. First, an email to “all involved individuals beginning on November 19, 2026.” The outreach will involve sending a notice by mail to involved individuals in North Dakota and Washington and notifying the regulators in those states. “We specify that this was a security incident involving GoFundMe Pro. We do not reference any nonprofit organization by name,” according to a statement from GoFundMe. A dedicated call center has been established to answer questions involved individuals may have beginning on November 19, 2026.”

Users of the GoFundMe Pro Salesforce integration should rotate all keys and credentials stored in a Salesforce environment. All Salesforce account activity should be reviewed, especially the Setup Audit Trail and Login History, for unfamiliar activity since October 14. Users should also review any users created or modified since that date. Salesforce also recommends reviewing Apex execution and API activity for anything unexpected.

The attacks appear to be coordinated. Said one nonprofit technologist speaking on background, “Keep going. The hole is way deeper.” Another told The NonProfit Times on background that in at least one case funds were diverted from a platform. The IP addresses of the bots are all outside the United States but could have been rerouted from the United States to mask the location.

“We have a lot of evidence to show these attacks are being run and they’re utilizing AI bots,” said Lauren Sheehan, president of SofterWare, While the attack has been contained, the DonorPerfect application is down and there are functionality issues on marketing pages and forms.

While impacted nonprofits can still accept online donations, some are not able to synch to their fundraising constituent relationship management systems, either because of the attack or for safety.

GoFundMe is posting updates on that issue here: status.gofundme.com. SofterWare is likewise keeping users up to date at https://status.donorperfect.com

“We sent an email to all of our clients letting them know that the investigation is ongoing and at this particular time we do not have a determination of who was impacted or what if anything was taken,” said Sheehan. “We just wanted to lean in and be transparent and get the word out. It’s kind of more eyes on things. If you see something, say something could be helpful.”

“We’ve got an investigation that is happening right now. We have identified the root cause of the incident … The initial access point was through our marketing website,” said Sheehan. “I can tell you that we were already investigating an issue when we got an email from another technology leader who was absolutely experiencing an attack from the same bad actor,” she said.

At this writing, the DonorPerfect (donorperfect.com) and SafeSave Payments (safesavepayments.com) public websites have been restored and are accessible. The DonorPerfect application itself remains temporarily unavailable.

A SofterWare official said that as a donation processing platform it is often the target of bad actors but have extensive systems and procedures in place to ensure that data and donations are secure. “What’s different this time is that an AI agent was rapidly adapting its approaches and discovering new methods to attack systems within seconds or minutes,” the official said on background.

“We’re sharing what we learned with other nonprofit technology providers to help them strengthen their own incident response,” according to an official at DonorPerfect. The system now “reviews security activity around the clock, summarizes unusual patterns for our response team, and helps staff check whether new protections interfere with legitimate activity, such as a donor making a gift or a nonprofit staff member signing in.  This system can’t change a single security setting without a named person’s approval.”

There are two parts that work together:

* A live security review document. A shared, continuously updated report that the response team, executives and engineers all read. It holds an hour-by-hour log of what was seen, a list of open actions (each with an owner and a status), a record of every protection that was added and why, and a separate view of sign-in activity on customer accounts. Anyone on the team can open it at any time and see where things stand, without waiting for a status meeting.

 * A live dashboard. Charts that show, hour by hour, how much of the traffic reaching SofterWare’s systems is ordinary use and how much is attack traffic, which protections are doing the blocking, whether donations are going through, and how sign-ins are behaving. Known incidents are marked on the timeline so the team can see cause and effect at a glance.

In its first 48 hours, it tracked a persistent attacker, confirmed every attack on the donation forms was stopped, and flagged within the hour when new protections blocked genuine visitors so that those settings could be immediately addressed by our staff, the official said.

“We’re sharing how we organize monitoring, flag findings for human review, and check whether protective measures disrupt fundraising activity with peers and competitors in the nonprofit technology space,” according to the official. Google Threat Intelligence Group’s November 16 report describes how attackers are using AI and how Google uses continuous monitoring in its own defenses. SofterWare developed its approach during the DonorPerfect response, and it reflects the same emphasis on spotting unusual activity quickly.”

This is not the last time something of this nature will occur, according to technology experts. “Recent security incidents affecting multiple fundraising platforms underscore the importance of protecting nonprofit and donor information,” said a GoFundMe official. “We’re responsible for safeguarding the information entrusted to us, and we take that responsibility seriously. Earning trust every day means being clear about what happened, taking action to address it, and keeping you informed as our work continues.”

The NonProfit Times will be updating the story as more information comes in.