ACA Group, a governance, risk and compliance adviser to financial firms, and the National Society of Compliance Professionals, a non-profit membership body for compliance staff in financial services (NSCP), reported on 12 November 2026 that 86% of surveyed firms formally use AI while 38% operate a formal testing program for AI outputs. Fintechbits checked ACA’s own release of 29 October 2025 for the second edition and found that 28% of firms tested or validated AI outputs, against 38% with a formal testing program in the 2026 survey. The third annual AI benchmarking survey drew 261 responses in September 2026 and shows adoption running well ahead of proof that the tools work.
What the AI Benchmarking Survey Measured in 2026
The 2026 AI Benchmarking Survey collected answers from 261 NSCP members and ACA clients, and ACA’s 2026 AI Benchmarking Report frames the theme as a move from AI adoption to assurance. Respondents were compliance and risk staff at hedge funds, broker-dealers, private equity firms, asset managers and alternative investment advisers, and the typical firm employed between 11 and 50 people while overseeing regulatory assets of $1 billion to $10 billion.
Adoption is high but mostly internal. ACA and NSCP found that 69% of firms confine AI to internal use cases, 18% also use it in client-facing tools, and roughly a quarter are trialling agentic AI. Enterprise or private generative AI products dominate, with 85% of firms using them against 18% on free public tools.
In the AI benchmarking survey, oversight is where the numbers weaken. Among surveyed firms, 60% apply human review to AI outputs on a continuing basis and 26% only when the need arises, but just 38% have formal testing. Some 81% of respondents say their oversight is matched to the risk of what the AI does, but only 51% are confident they could hand over records of that oversight quickly. The top concerns are information security at 68%, data privacy at 59% and cybersecurity at 55%, no technical control is used by more than 36% of firms, and 18% have an incident response plan covering AI-specific disruptions.
How the 2026 Results Compare With the 2025 Edition
The 2025 edition of the AI benchmarking survey, published by ACA on 29 October 2025, put formal AI use at 71% of firms, a 26-point rise from 2024. That implies roughly 45% in 2024, so formal use has almost doubled in two surveys. The 2026 figure of 86% adds a further 15 points, and the pace of adoption is slowing as the base grows.
Governance has moved too. ACA reports AI governance committees at 66% of adopting firms, up from 48%, and enhanced due diligence on AI vendors at 55%, up from 43%. Another 24% of respondents are building formal testing programs, and testing was the most common priority named by compliance leaders for the following 24 months.
One caution applies to the testing comparison. The 2025 release described firms that test or validate outputs, while the 2026 release describes a formal program, and ACA does not say the question was unchanged. The 10-point gain is therefore an indicative reading, not a like-for-like trend line, and Fintechbits would not build a forecast on it.
SEC Exam Priorities and AI Washing Enforcement Frame the Gap
The documentation figure matters because the SEC examines advisers on exactly this point. The Division of Examinations’ fiscal year 2027 examination priorities, released on 23 October 2026, say the Division will prioritize examinations of policies and procedures for monitoring what the document calls super intelligence, including its use in fraud prevention and detection, back-office operations, anti-money laundering and Regulation Best Interest compliance, and trading. Fintechbits covered the monitoring side of that AML use in AI in AML Compliance.
Enforcement shows what loose AI claims cost. On 18 March 2024 the SEC settled charges against investment advisers Delphia (USA) Inc. and Global Predictions Inc. for false and misleading statements about their use of AI, with civil penalties of $225,000 and $175,000, or $400,000 in total. Neither firm admitted or denied the findings. ACA President Carlo di Florio notes that the SEC “has not adopted a rule specifically governing AI use”, but says fiduciary duty, recordkeeping and vendor oversight obligations still apply to firms using it.
Fintechbits analysis
Why Formal AI Testing Is the Compliance Gap to Close
The number to watch in the AI benchmarking survey is 51%, not 86%. Adoption among 261 respondents says little about the wider market, but the spread between belief and evidence is hard to explain away: 81% of respondents think oversight fits the risk, 51% could show proof quickly, and 38% test outputs formally. That leaves 49% of respondents without confidence in their own paper trail, which is the exact thing an examiner asks for.
Two cautions apply. ACA sells compliance advisory services and an AI product called Encore AI, so a report that finds governance gaps also describes part of its sales pipeline. The respondents are also self-selected, and they are asset managers, private fund advisers and broker-dealers rather than banks or fintech lenders, so the percentages do not carry over to payments or lending firms.
The direction is clear. Regulators treat unverifiable AI claims as a supervision and marketing problem. Smaller teams without a full compliance function face the same test, and Fintechbits asked industry leaders about fractional compliance help. The 10-point gain in testing is real progress, but it is too slow for firms that already put AI in front of clients.
Frequently asked questions
What is the ACA NSCP AI benchmarking survey?
The AI benchmarking survey is an annual study by ACA Group and the National Society of Compliance Professionals on how financial services firms adopt and govern AI. The 2026 edition is the third and gathered 261 responses in September 2026.
How many firms run formal testing in the AI benchmarking survey?
The 2026 AI benchmarking survey found that 38% of firms operate formal testing for AI outputs, while another 24% are developing such programs. Only 51% of firms said they could hand over records of their AI oversight quickly.
Does the SEC have a rule on AI use by investment advisers?
ACA President Carlo di Florio says the SEC has no AI-specific rule, so existing duties on fiduciary conduct, recordkeeping and vendor oversight apply. The SEC Division of Examinations lists monitoring of the technology in its fiscal year 2027 priorities, released on 23 October 2026, and the SEC fined Delphia and Global Predictions $400,000 in total in March 2024.
What to Watch Next
The 12 November 2026 AI benchmarking survey release leaves three open points. The first is whether ACA and NSCP repeat the testing question in a fourth edition with the same wording, which would turn the 38% figure into a true trend line. The second is whether SEC examinations under the fiscal year 2027 priorities, released on 23 October 2026, produce findings or enforcement cases that cite weak AI testing or missing oversight records. The third is whether the 18% of firms with an AI-specific incident response plan rises, since preparation for AI disruptions is the weakest control in the survey.



