AI in Finance #002 — Do We Need AI to Solve This Problem?
Originally published in Weekend Reads on LinkedIn · Subscribe
Regulatory and compliance adoption in AI is one of the biggest hurdles organizations are facing. Here are a few incidents where organizations had to face legal action when their AI based solutions failed to act responsibly:
Deloitte Australia — AI-hallucinated government report (Oct 2025). A corrected version was later published in which certain sources were deleted, including fabricated academic papers and a made-up quote misattributed to a federal court judge — in an assurance report by the firm whose entire job is assurance.
Wells Fargo — "Enhanced Credit Score" mortgage algorithm (verdict pending). A consolidated racial-discrimination class action over an automated underwriting system that plaintiffs say disproportionately denied 100,000+ minority refinance applicants.
Louis v. SafeRent (settled ~$2.28M, Nov 2024). Plaintiffs alleged the algorithm behind SafeRent Scores relied on credit-scoring data reflecting historic racial disparities and ignored housing vouchers, disproportionately denying Black and Hispanic applicants; the court found they had plausibly alleged disparate impact.
In the research I did for this article, I found 50-plus reported incidents in the past three years where AI systems hallucinated or fabricated content — and these are only the ones that made the news. Public trackers like the OECD AI Incidents Monitor, which draws on over 150,000 news articles a day, show incidents rising sharply since 2022, so the true count is far higher. The Future Society notes that the number of monthly AI-related incidents and hazards tracked by the OECD surged by 50 percent over the last six months. I'm sure there are many more that went unnoticed.
Organizations are in a mad race to increase productivity. Clients are asking how AI can help them grow, and organizations are responding with solutions that are fast and efficient — but are they reliable?
The bigger question that should be asked before building any AI solution: do we need AI to solve this problem? Many problems can be solved without it — not every solution has to be built on AI. It's a judgment call, made on the merits of each approach.
My take: use AI to build systems that are more deterministic and less probabilistic. Use AI as an accelerator to build the systems — don't make AI the core of your solutions.
Is it possible to build solutions that are more reliable and compliant? I will answer this question in articles to come.
More on this next week - Same time, Saturday.
References
Deloitte Australia — AI-hallucinated government report
AP via Yahoo: yahoo.com/news/articles/deloitte-partially-refund-australian-government-070855665.html
CFO Dive: cfodive.com/news/deloitte-refunds-60k-report-ai-errors-australian-government-accounting/803321/
OECD AI Incidents Monitor entry: oecd.ai/en/incidents/2025-10-05-be45
Wells Fargo — "Enhanced Credit Score" mortgage algorithm (pending)
Civil Rights Litigation Clearinghouse (docket, In re Wells Fargo Mortgage Discrimination Litigation, 3:22-cv-00990, N.D. Cal.): clearinghouse.net/case/45306/
Bloomberg Law: news.bloomberglaw.com/banking-law/wells-fargo-racial-disparity-case-heads-to-class-action-decision
Louis v. SafeRent Solutions (settled ~$2.28M, Nov 2024)
Quinn Emanuel analysis (confirms the disparate-impact finding and settlement): quinnemanuel.com/the-firm/publications/when-machines-discriminate-the-rise-of-ai-bias-lawsuits/
Class counsel, Cohen Milstein (grab the specific Louis v. SafeRent page before publishing): cohenmilstein.com
The Future Society, citing the OECD Monitor: thefuturesociety.org/us-ai-incident-response/
OECD Monitor itself (methodology + the 150,000-articles-a-day figure): oecd.ai/en/incidents and oecd.ai/en/incidents-methodology
All sources verified as of 21 August 2026. Wells Fargo is pending and allegations are unproven; case statuses evolve — verify current posture before relying.
AI in Finance
Have thoughts on this?
Let's continue the conversation on LinkedIn.
Get new posts by email
Occasional notes on the BA → AI PM journey. No spam.