A privacy program might look roughly the same today as it did two years ago, the same information collected, the same categories flagged as sensitive at intake, the same methods of obtaining consent, etc., but it might have fallen out of step with the law because legal definitions of what counts as sensitive personal information (SPI) have started to move.
This year, three states have done just that, with both Maryland and Connecticut expanding what qualifies as SPI under their respective privacy laws, while New Jersey’s new data broker law, signed at the end of June 2026, imposes a hard restriction by prohibiting the sale of SPI outright.
This is part of a broader pattern where the list of categories that trigger heightened privacy obligations keeps growing, and in at least one state, sensitivity can attach to personal information a company never directly collected. Understanding why, and what it changes inside a privacy program, matters for the years ahead, as this trend is likely to maintain its trajectory.
Table of Contents
Why is sensitive data expanding in the US?
When most state privacy laws were first written, the definition of SPI borrowed heavily from a fairly narrow, largely GDPR-influenced template:
- race or ethnicity
- religious beliefs
- health conditions
- sexual orientation
- citizenship or immigration status
- biometric or genetic information
- geolocation
That template worked when most information fell cleanly into a sensitive or personal information bucket, but two forces are now pulling it apart and making the landscape a bit more complex to navigate.
New Technology is Creating New Types of Data
Emerging technology has created entirely new categories of information that simply did not exist before. Neural data is a clear example of this. For instance, technology increasingly used in meditation apps and brain-computer interfaces captures central nervous system activity. This is an example of neural data, a new category of sensitive personal information that is starting to pop up in new privacy laws and amendments to existing privacy laws.
But there is a little more complexity here, and in this case, peripheral neural data is also starting to appear across the privacy landscape and can be found, for example, when a wristband reads electrical signals in forearm muscles to translate hand gestures into computer commands. Since the signal originates in a nerve outside the brain and spinal cord, this type of information counts as peripheral neural data.
Connecticut’s SB 1295 defines neural data narrowly, central nervous system activity only, while Colorado and California take the broader position that peripheral nervous system activity (activity in nerves that extend out from the brain and spine, such as a wrist-worn device that uses muscle signals to let someone control another device) counts too – a good illustration that state privacy laws are not converging on a single understanding of SPI, even as the lists keep growing.
And it’s not just neural data. The rise of direct-to-consumer genetic testing, wearable devices, and facial recognition technology, among others, all add to an increasingly long list of personal information being covered by SPI classification.
What is Now Classed as Sensitive Information?
Oregon’s privacy law in particular shows how far one amendment can reach once a state starts naming categories explicitly by adding status as a victim of a crime as a defined category. This carves out protections for information revealing whether someone has experienced stalking, domestic violence, or assault and recognizes that this type of information carries its own heightened risk.
Oregon, along with Connecticut, Delaware, New Jersey, and Maryland, have also named transgender or nonbinary status as another example of the new categories of sensitive information being found under state laws. Naming new instances of sensitive information explicitly removes ambiguity about coverage and highlights where legislative attention is heading next.
Taken together, these recent amendments – and those similar – have added a host of different types of information to the list that companies need to watch across multiple states. Even where two states use the same label, the scope can differ. Take New Jersey, for example; it defines financial information broadly, while Connecticut ties its version specifically to an account number, login credentials, or a card number combined with the code or password that grants access.
The more consequential shift, though, is not which new fields made the list, but where sensitivity can now attach itself. Maryland’s HB 711 amends the Maryland Online Data Privacy Act (MODPA) so that the information an organization infers from other personal information can be classified as SPI if that inference indicates a category already defined as SPI. So a company does not necessarily need to collect a field labeled “health condition” to process SPI under Maryland law. A conclusion drawn from browsing behavior, purchase history, or engagement patterns can count on its own.
This is not hypothetical either. When HB 711 was moving through the legislature, one privacy attorney flagged pharmaceutical advertising as a real example. Because Maryland already bans the sale of SPI outright, a company relying on advertising to an audience using inferred health conditions could move from believing it collects no health data to unknowingly selling SPI, without changing a single source. Not everyone agrees on how far this reaches, as some see a real shift for health-adjacent advertising while others say inferred data was already captured under the prior definition.
Connecticut’s amendment tells a related but distinct story. Beyond adding categories, SB 1295 removes the processing volume thresholds that previously determined whether a company had to comply with the Connecticut Data Privacy Act (CTDPA) on SPI. That means that any entity processing Connecticut residents’ SPI is now in scope regardless of volume, so a company that never came close to the original thresholds can now be squarely inside its scope. Vermont’s new privacy law (which comes into force on January 1, 2028) comes close to this approach without going all the way, since it applies to businesses processing the sensitive data of as few as 3,000 Vermont residents, a floor low enough to catch many companies that would sit well below a general consumer threshold. This represents a more structurally significant change for privacy teams than the expanded definition of SPI itself, since the trigger is moving from a question of volume toward a question of information collected.
New Jersey’s recent legislative activity illustrates a third variation, an expanded definition paired with an outright restriction. Its new data broker law, signed at the end of June 2026 and effective immediately, prohibits data brokers and data collectors from selling SPI, and separately amends the state’s comprehensive privacy law to prohibit any entity, regardless of size, from doing the same, with penalties up to $50,000 per record. However, the state has signaled it will hold off on enforcement while it fixes defects in the bill.
What impact does this have on privacy programs?
Putting the developments in those three states together alone highlights a clear pattern where most privacy programs assumed SPI could be identified at collection, by tagging known sensitive fields and treating everything else as personal information. And that was true when the list was shorter and tied to what a company directly collected. Less so now, however, because the list of SPI is growing and because, at least in Maryland, SPI classification can be attached downstream from inferred sources rather than in anything a company actually gathered directly from an individual.
While focusing on collection and classification of SPI remains squarely in focus, it is important to consider other key aspects of a privacy program. This includes areas such as:
- Consent built around a general privacy notice that may not satisfy new standards
- Risk assessments that were reviewed once and filed away
- Privacy rights processes that aren’t scoped to cover where SPI has been inferred
- Previously negotiated vendor agreements that may not obligate a third-party to supply what organizations need for a risk assessment
- Security controls that were calibrated to a more limited list of SPI and may not reach newly regulated information like government IDs or neural data
The newly established (albeit ever-changing) landscape for SPI doesn’t mean that company privacy programs need a complete overhaul overnight; however, companies should focus on the areas that are most likely to represent a blind spot where SPI can exist unnoticed.
What do companies need to do about it?
Collection and inventory
The starting point is extending the data inventory beyond directly collected fields to include inferences derived from personal information, not just where health or biometric data lives, but where a process might generate a conclusion that indicates a protected category, even if no source field was ever labeled as sensitive. This is also the place to account for tracking technologies like cookies and pixels, a common source of the behavioral data that feeds this kind of inference. Rather than deciding whether a given pixel implementation on its own constitutes the processing of SPI, include tracking technologies in a data inventory like any other data source. A scanning tool like Boltive, Observepoint, and Privado can help surface what is actually running on a site or app.
Risk assessments
Connecticut, Maryland, and New Jersey require organizations to conduct and document what their statutes specifically call a data protection assessment when processing SPI, and they are far from alone, since many other states with comprehensive privacy laws require the same thing.
Furthermore, California requires what its regulations call a privacy risk assessment before processing sensitive personal information, an obligation that has applied since January 1, 2026. Most companies do not need to build this process from scratch, because an existing privacy impact assessment or risk assessment, including one already built to meet the CCPA requirement, can usually be adapted to satisfy each state, as long as the output addresses what each state’s law asks for. The bigger change is treating that assessment as a living process rather than a once-a-year exercise, since processing activities, including ones that never needed a risk assessment before, can quickly get pulled into scope through new laws and amendments. Reviewing existing activities whenever a state amends its definition, rather than waiting for the next scheduled cycle, is the practical fix.
Consent
Newly regulated types of SPI need consent mechanisms that can cover the right types of information, and the mechanism needs to match the specific legal standard rather than assuming consent is consent everywhere. In states that still use opt-in consent for SPI, a well-scoped consent flow remains an important control, and it is worth checking that each newly added category, including data inferred from other information, is actually captured by it.
Maryland limits collection and processing to what is strictly necessary for a product or service the consumer requested, regardless of consent. And both Maryland and New Jersey prohibit the sale of SPI with no consent exception, so a consent banner cannot be relied on to make either of those activities lawful, and any sale of this data needs to be identified and stopped rather than covered with a checkbox.
Privacy rights
Privacy rights fulfillment processes need to account for inferences, not just collected fields. Connecticut’s amended law gives consumers the right to access inferences drawn from their personal data, so a process built around querying a data source for stored fields will not satisfy that right if a company generates inferences that never get documented in a retrievable way.
Security
Newly defined types of SPI deserve the same technical safeguards as those that are already treated as high risk. Government IDs, financial account credentials, and neural data warrant the same encryption, access controls, and monitoring already applied to health or biometric data. The task is usually less about designing new controls and more about confirming these categories have actually been brought inside the scope of existing security programs.
Third-party agreements
Vendor contracts need revisiting to reflect the new restrictions and documentation obligations, and that challenge shows up in two different places. Some amendments now require third parties to supply organizations with the information necessary to complete a risk assessment. Others attach an outright restriction on selling SPI, which raises a more basic question for every vendor relationship touching SPI: does the agreement actually prohibit what the law now prohibits, or does it rely on older language that predates these changes?
A Final Word on SPI Under US State Privacy Law
The list of what counts as SPI is not going to stop growing, and many states moving to treat certain inferred data as SPI suggests future amendments may keep expanding not just which types of information count, but where SPI can exist. A program that treats this as a one-time update, tagging a few new fields and calling it done, is likely to have the same conversation again next year. A program that treats it as an ongoing discipline, revisiting its inventory, risk assessments, consent, privacy rights process, security controls, and vendor agreements whenever a state adds new requirements, will be in a much better position no matter what comes next.
At Red Clover Advisors, we help organizations of all sizes navigate complex privacy challenges and can help guide you through what your data inventory, risk assessments, and vendor agreements need to look like as SPI keeps expanding. If need help aligning your privacy program to the ever-changing privacy landscape, schedule a consultation today.
State Privacy Laws Comparison Guide
Stay ahead of state privacy laws with our guide—clear definitions, key dates, and crucial compliance tips!