Sei AI

Sei helps customers find regulatory violations in text and images. Oh and look at how the model catches issues in Spanish as well!! #artificialintelligence #ai #compliance #consumerprotection

?? ?????? ?????????????????????? ???? ????2024: ?? ????????????-???????????????? $8.2 ?????????????? ???? ?????????????????? ???????????????? ?? The SEC filed 583 enforcement actions in FY2024, securing an unprecedented $8.2 billion in financial remedies. While enforcement actions declined 26% from last year, the Commission maintained a strong focus on emerging threats, including artificial intelligence misuse and crypto fraud. ?????? ????????????????????: ?? ????-?????????????? ??????????????: Firms faced charges for false claims about AI's role in investment decisions and performance. ?? ????????????-???????????????? ??????????????????: $4.5B+ financial remedies in the Terraform Labs fraud case alone. ?? ?????????????????????????? & ????????????????????: Over $600M in penalties for record-keeping violations and proactive measures addressing marketing rule breaches. ?? ?????????????????????????? ????????????????????: $255M in awards and a record 45,130 tips received. ?? ?????????????????? ??????????????: Companies practicing self-reporting and remediation saw reduced or no penalties, encouraging a culture of compliance. As SEC Chair Gary Gensler sums up: "The Division of Enforcement is a steadfast cop on the beat, promoting the integrity of our capital markets to benefit investors and issuers alike." Read more about the SEC’s Enforcement Results for Fiscal Year 2024 (link in comments) ?? #Compliance #InvestorProtection #Finance

Of course, my weekend read was the 2024 AI Benchmarking Survey by NSCP and ACA Aponix! It’s a fascinating survey on how finance-related firms are using AI tools and tackling compliance practices (link in comments). Here are some highlights: 1.? ?Only <8% use customer-facing AI tools ??, while 37% use it internally. High potential for more adoption! 2.? ?50% of firms use Private/Enterprise Gen AI other than ChatGPT ??. Pleasantly surprised that it’s this high! 3.? ?Just 31% of companies using AI tools are doing so for compliance ???. Huge potential if you’re building in this space like us at Sei AI ??. 4.? ?68% lack a framework to manage AI risk, and 61% don’t have a dedicated team to develop/maintain AI tools. Only 18% have a formal testing program for AI tools ??. 5.? ?Most firms (56%) don’t use AI for creating marketing material. Among those who do, 60% said that marketing, legal, or compliance teams review all AI-generated material. Finance is definitely lagging behind other industries here. The most interesting data point for me was that 60% of companies see regulatory risk as the top concern for using AI tools ??. This seems closely tied to another stat in the report: 90% of companies do not market their use of AI ??.

?????? ?????????? ?????????????????? ?????????????? ???????????? ???????? ?????????????? ?????? ???????? ?????? ?????????????????? ?????????????????? The Federal Trade Commission (FTC) has accused Dave of misleading marketing, charging undisclosed fees, and collecting unauthorized ‘tips’ from users on its app. Here’s a summary of the charges: ?? ???????????????????? ???????? ????????????????: Dave often fails to deliver the ‘up to $500’ advances it advertises, with many users only receiving a fraction of the amount (screenshot below) ?? ?????????????????????? ????????: Consumers were charged an ‘Express Fee’ for instant access, ranging from $3 to $25, which wasn't disclosed until after account access was granted. ?? ???????????????????????? ‘????????’:?Dave imposed a surprise 15% fee on advances, disguised as a ‘tip’, with many consumers unaware of the charge or how to avoid it (screenshot below). ??? ?????????????????? ???????????????? ????????????: Dave's interface misled users into believing their ‘tip’ was funding meals for the needy, while in reality, it only donated a small fraction. The FTC’s complaint alleges that Dave’s actions violate consumer protection laws and calls for stronger enforcement to protect vulnerable consumers. The majority of its customers are described by it as “financially vulnerable” or “financially coping,” including those live paycheck-to-paycheck, who have minimal savings, and who overdraft their bank accounts frequently. #ConsumerProtection #FTC #FinTech #OnlineLoans #DeceptiveMarketing #FinancialTransparency

A second-order effect of AI tools - businesses are increasingly looking for OaaS (Outcome as a Service) rather than SaaS (Software as a service). For the last few decades - SaaS companies have pitched their products as productivity enhancers for average employees. The business logic was that your employee who earns $50,000/year is now 25% more productive. You can purchase our software for $25,000 annually, hire 8 employees instead of 10, and save $75,000. But LLMs can give highly-productive employees superpowers. This means the employee who is earning $200,000/year is now 100% more productive. But in the real world - salaries and workloads aren’t that flexible. Employers can’t double employees’ scope of work or reduce salaries by 50% overnight.? ? The solution that companies are looking for is ‘buying work, not just software’. You have an AI tool for auditing financial documents? Bundle it with a qualified human who is in the loop. An AI code generator? Add a developer who is qualified to review and sign off on it. The latter isn’t even hypothetical - Google just announced last week that 25% of its code is AI-generated - with employees only reviewing and accepting it. My guess is that you’ll probably see a new kind of AI-first marketplace opening up on the lines of Upwork or Fiverr. With highly-qualified experts who leverage LLMs to produce fast, cheap high-quality (and high-margin) output.

There’s a lot of talk about ‘selling shovels during the gold rush’ during the current AI boom. But unlike the actual California gold rush of the 1800s - the quality of the shovels matters a great lot now ?? For example, financial institutions we speak to are exploring using AI chatbots for customer support - but are mildly terrified about the legal ramifications of a bot hallucinating. So a common ask we get is - can you QA our bots and make them fool-proof? The recent enforcement action on VyStar has brought the focus on adequately testing your systems before rolling them out. Needless to say - this is easier said than done. An extensive QA for chatbots would comprise of: ??? Accuracy Checks: Ensuring that generated responses align with bank policies and provide accurate information. ??? Compliance & Security: Confirming adherence to financial regulations and safeguarding sensitive data. ?? Bias & Fairness Monitoring: Identifying and mitigating any unintentional biases to promote inclusivity and fair service. ?? Performance Optimization: Maintaining consistent performance even during high traffic. At Sei AI - we’re currently building a comprehensive pre and post-deployment QA solution for banking-related chatbots. If you’re interested in knowing more - DM me! Would love to chat.

$89.8 ?????????????? ???? ??????????????????: ???????? ???????????? ???????? ???? ?????????? ?????? ?????????????? ?????????? ?????? ???????????????????? ???????????????? On 23rd October, the Consumer Financial Protection Bureau (CFPB) ordered Goldman Sachs to pay at least $19.8M in redress along with a $45M civil penalty, and Apple to pay $25M civil penalty.?It also banned Goldman Sachs from launching a new credit card till it provides a credible plan to comply with consumer regulations. This case pertains to the Apple card that was launched in partnership with Goldman Sachs in 2019. Here a quick summary on why the CFPB ruled that they harmed consumers: 1. Dispute Mishandling: Apple failed to send tens of thousands of consumer disputes to Goldman Sachs due to a misleading UI/UX feature on the Wallet app and the Messages app, leading to unresolved cases and customer frustration. 2. Ignored Federal Requirements: Goldman Sachs didn’t follow federal guidelines for investigating disputes within 30 days and sending a resolution with 90 days. This resulted in delays, unresolved charges, and incorrect negative credit reports. 3. Inadequate Payment Transparency: Apple’s deceptive marketing materials and illegal conduct caused consumers to pay interest on purchases of iPhones and other Apple devices with an Apple Card that they believed would be covered by an interest-free payment plan. 4. Technological Gaps at Launch: Despite warnings about readiness issues, Apple and Goldman launched Apple Card without a reliable dispute system, meaning consumers faced long waits to get money back for disputed charges. Follow Sei AI for the latest on AI-related marketing and compliance updates, news, insights and more.

It's long overdue, but the FTC finally came down (hard) on the manipulation of customer reviews. From humans and AI both. Here’s a gist (full report in comments): ? No more fake or AI-generated reviews ?? Prohibits paying for positive or negative reviews ???♂? Cracks down on insider reviews without clear disclosure ?? Bans suppressing negative reviews through threats or intimidation ?? Stops misuse of fake social media influence (bots, fake followers) Customer reviews are the cornerstone of growing a business organically - so it’s obvious that there is a massive incentive to manipulate them. At Sei - we’ve been working on this (admittedly complex) problem for a while now. Our agents can flag AI-generated testimonials by comparing their tone and consistency to real ones. They also use other subtle signals. For eg - does the review accurately describe the product/service being reviewed? As bots learn to sound more and more ‘human’ - AI compliance teams will always be in a race to stay one step ahead. The FTC’s ‘final rule’ helps enable them further - protecting consumers as well as honest businesses.

Dark patterns have been a huge area of concern for the FTC over the last few years. This is why they officially adopted the ‘Click to Cancel’ rule yesterday, making it easier for users to cancel their online subscriptions (should be as easy as subscribing in the first place) This rule creates stringent civil penalty liabilities for: 1.? ?Any misrepresentation related to the marketing of a product or service 2.? ?Failure to disclose all terms and conditions before obtaining billing information? 3.? ?Failure to obtain express informed consent before charging users 4.? ?Failure to provide a simple mechanism for canceling a subscription At Sei AI - our AI agents can both check if relevant disclosures are present and flag any unsubstantiated claims that can mislead consumers. The icing on the cake is our ability to classify and escalate customer complaints, so companies can proactively monitor and prevent complaints from bottling up. With other similar cases (like the Adobe lawsuit) currently in court - expect many more regulations related to dark patterns in the near future.

A recent Thomson Reuters report states that 67% of compliance teams expect an increase in compliance costs in the next 12 months, but only 50% expected an increase in their compliance budget in the same time period. As AI compliance agents grow in popularity, here’s how they help navigate increasing regulatory scrutiny while also balancing cost pressures: ???? Respondents said that the 3 most important qualities they look for in a compliance officer were subject matter expertise, attention to detail and integrity. AI models are far less prone to bias compared to humans, and can detect minute issues that manual detection might overlook. ? Nearly 30% of respondents said that they spend more than 8 hours a week just tracking the latest in regulatory developments. AI can reduce this substantially by applying the latest rules and regulations to compliance detection models. ?? Only 38% of survey respondents outsourced compliance requirements. Most teams would prefer to improve in-house productivity and skillsets instead of outsourcing - AI helps supercharge existing compliance teams. Sei AI is building powerful AI compliance agents that enable fast-growing teams to scale in a compliant, cost-effective way. Follow us to know more.