Who gets hired
Your AI can’t disfavor candidates on protected class: race, sex, age, disability. Outcomes are scrutinized, intent or not.
As AI takes on more of your business, its mistakes can become your company’s legal problems*. Privlex brings continuous oversight and qualified counsel together, so you can find out where your systems are going wrong and address it through a process designed to support attorney-client privilege.
Every week in the news there’s another story about an AI agent going rogue. But your system doesn’t have to make headlines to create legal exposure. A dependable workflow can keep doing exactly what it was built to do while making hiring decisions, speaking to customers, or handling personal information in ways that put the company behind it at risk.
That’s what makes this a business and a technology problem. You need to understand what your AI is actually doing, where that behavior crosses a legal line, and what to do when it does. The examples below show how ordinary AI activity intersects with the obligations your business already has.
Your AI can’t disfavor candidates on protected class: race, sex, age, disability. Outcomes are scrutinized, intent or not.
Your AI’s lending calls carry liability for protected-class outcomes, plus a duty to explain denials.
Your AI steering ads or scoring tenants can’t disfavor protected classes; disparate impact alone is liability.
Your AI in patient care can’t yield discriminatory outcomes; HHS now reaches decision-support tools directly.
Your AI rating risk can’t produce unfair discrimination, and states now make you test and govern the models.
When your AI scores someone, they can demand the reasons and dispute the result; private suits attach.
Your AI can’t use or infer genetic or family-health data in an employment call; flat ban.
Take federal money and your AI can’t discriminate by race or national origin; one hook into health, schools, benefits.
These duties land state by state, and no two regimes agree: CA, NYC LL144, IL, TX, and counting.
Every false thing your AI tells a customer; FTC, 50 state AGs, and under state law the customer.
How your AI pressures, steers, or exploits a customer; actionable even when nothing it said was false.
Your AI inside a financial product can be unfair, deceptive, or abusive; CFPB and state regulators bring these.
When your AI asserts a falsehood about a real person, tort liability attaches; no statute, no cap.
If your AI conjures someone’s face, name, or voice to sell, that's their property and their lawsuit.
Reviews or testimonials your AI fabricates are deceptive per se, and the FTC penalizes them.
Your AI dialed it: $500–$1,500 a message, and consumers can sue.
How your AI contacts people about debts (harassment, falsehoods, timing) all still bind; private right of action.
Your AI’s marketing mail must honor unsubscribes and identify the sender; liability accrues per message.
If your AI profiles or targets children, consent rules are strict and FTC penalties steep.
If your AI reads a face or voiceprint, each person gets a private suit and statutory damages.
PHI moving through your AI’s training or inference triggers privacy, security, and breach duties; OCR enforces.
When your AI monitors workers, the NLRB can treat it as interfering with protected organizing.
Your AI’s chat, scoring, and scheduling must work for people with disabilities; DOJ-readable, private suits.
For companies trying to use AI responsibly, the incentives are uncomfortable. Looking closely at a model or workflow can uncover an issue that now needs to be remediated, along with a record of what the company knew. That makes it tempting to postpone scrutiny, especially when the system appears to be working and the business is under pressure to keep moving.
But not looking doesn’t resolve the underlying problem or absolve a company of responsibility for what its systems do. The better answer is to give diligence a beneficial structure: involve qualified counsel from the start, so that when an issue arises, the company can seek legal advice and have counsel guide the remediation under privilege where the applicable requirements are met.
That is why Privlex loops counsel into oversight. Finding a potential problem becomes the beginning of an informed response, with legal judgment built into how the issue is assessed and addressed.
Privlex measures what your AI actually does using legal classifiers tailored to your business. Counsel helps establish the requirements that matter, and our monitoring turns activity across your systems into structured findings that can be reviewed in that legal context.
When a finding needs attention, qualified counsel can assess what it means and advise how to respond. The point is to connect discovery with remediation through a counsel-led process, rather than leave your team with another dashboard of risks and no clear path to resolving them.
For the majority of customers, you don’t even need to install new software, and your data isn’t copied out. We begin with the systems you use, the decisions they make, and the information they handle. Working with counsel, we map that activity to the relevant legal requirements and determine where monitoring will be most useful. Those requirements then become the basis for classifiers that evaluate real outcomes as your business operates.
As findings emerge, counsel reviews the issues that need legal judgment and advises the next steps. Your team can focus its remediation on what matters, while the oversight continues as your models and workflows change.
We map your systems and data to the legal obligations relevant to your business, with counsel helping prioritize the risks that deserve attention.
We translate those requirements into custom legal classifiers, with counsel informing the thresholds and context for reviewing findings.
We monitor outcomes continuously and surface potential issues to counsel, who can assess their significance and advise remediation.
The companies willing to examine their AI should be better equipped to use it responsibly. Privlex gives that willingness a practical foundation: continuous visibility into what the systems are doing, qualified counsel to interpret the issues, and guidance to help the business act on what it learns.
Teams can bring concrete findings into legal review, giving counsel the context to advise what needs to change and helping the business move toward a considered decision.
Oversight follows the outcomes of your AI across the workflows that matter, helping you identify concerns that a review of the model alone may miss.
As models and workflows change, monitoring continues to evaluate their outcomes against the requirements counsel has helped establish for your business.
We’ve built enterprise systems for regulated enterprises, represented Fortune 500 companies, advised the White House and federal regulators, and led delivery in defense and intelligence environments.
The Modern Digital Supply Chain
A technology company can bring a product to market without owning the infrastructure it runs on, developing the models it uses, or operating the services that connect it to customers. From Buick’s wheel to AI voice agents, understanding how those contributions fit together—and what responsibility accompanies them—is part of understanding your own product.
By Joe Ewing, Co-Founder and CTOTell us how your business is using AI and where you need a clearer view of the risk. We’ll discuss how Privlex can bring ongoing oversight and qualified counsel into that work, so your team has a practical way to act on what it discovers.
hello@privlex.com