FDA's TEMPO Pilot Lets AI Devices Reach Patients Pre-Clearance

The FDA has accepted four devices, including generative AI products from Cadence and Limbic, into its TEMPO pilot, a program that permits digital health companies to put products in patients' hands before receiving marketing authorization.

The Food and Drug Administration has admitted four devices into a pilot program that does something the agency has historically been reluctant to do: let a medical device reach patients before it has been cleared or approved for marketing. Among the accepted products are generative artificial intelligence tools from Cadence and Limbic, according to STAT News.

The program is called TEMPO. Its premise is that generative AI software does not behave like a hip implant or an infusion pump. It changes. It is retrained. Its outputs are probabilistic rather than deterministic, which means the traditional device review — freeze the product, test the frozen version, authorize that exact configuration — fits it badly. TEMPO offers an alternative sequence: get the product in front of patients under agency oversight, gather evidence from real use, and settle the authorization question later.

What changes when authorization stops being the starting gun

For a digital health developer, marketing authorization has always been the gate that separates spending from earning. Everything before it is cost: engineering, clinical evidence, regulatory consultants, and the long quiet stretch while a submission sits in review. Everything after it is revenue, or at least the possibility of revenue.

TEMPO moves that gate. A company accepted into the pilot can put a product into clinical use — and, in principle, begin building the commercial and evidentiary record that supports it — while the formal authorization question is still open. That compresses the gap between a finished product and a paying user, which is the single largest drag on the economics of small digital health companies.

It does not, however, remove regulatory risk. It relocates it. Under the conventional path, a company learns whether it has a marketable product before patients ever touch it. Under TEMPO, patients touch it first and the verdict arrives afterward. A product that has been deployed, integrated into clinical workflows, and used to make care decisions is a harder thing to unwind if the agency ultimately says no.

Why generative AI forced the issue

Generative models are the reason a pathway like this exists at all. A conventional software-as-a-medical-device submission describes a fixed algorithm with a fixed input-output relationship. Regulators can test it against a validation dataset and reason about how it will behave in the field.

A large language model applied to, say, clinical documentation or a mental health intake conversation offers no such guarantee. The same prompt can produce different text. The model may be updated. Performance can drift as the population using it shifts. Reviewers cannot certify a frozen artifact and be confident the certification still describes the product a year later.

The regulatory options were narrow. The agency could hold the line and effectively keep generative AI out of authorized clinical use, watching adoption happen anyway in the unregulated space around it — wellness apps, scribes marketed as administrative rather than clinical, tools deployed inside health systems under enterprise contracts. Or it could build a supervised channel and collect data from real deployment. TEMPO is the second choice.

Cadence, Limbic and the shape of the accepted cohort

Cadence and Limbic are two of the four devices recently accepted. Both are private companies, which is worth stating plainly: there is no listed equity here for investors to express a view on directly, and none of the four is a public-market instrument.

That is precisely why the pilot matters to the wider sector. The list of accepted devices functions as a signal about what kinds of generative AI applications the FDA considers appropriate for supervised early release. Every other digital health developer building in adjacent territory will read the cohort composition as guidance about its own odds — which clinical settings the agency is comfortable with, what evidence generation commitments come attached, and how much post-market monitoring the arrangement demands.

What venture and strategic investors will be watching

Regulatory timelines have shaped digital health valuations for a decade. A company with a clearance is worth materially more than the same company without one, and the discount applied to pre-clearance assets reflects both the wait and the chance the answer is no.

A pathway that lets revenue begin before that answer arrives changes the arithmetic of a funding round. It shortens the runway a company needs to reach commercial traction and gives acquirers something they have rarely been able to buy: a generative AI clinical product with actual usage data behind it rather than a validation study and a submission number.

The countervailing point is that early release invites early scrutiny. Adverse events, model failures, or clinician complaints will surface in a live deployment, and they will surface publicly. A pilot participant that stumbles does not merely lose its own authorization case — it supplies the argument against the entire pathway.

The wider market backdrop

The news lands in a market that has been paying for AI exposure with enthusiasm elsewhere in the economy but has treated healthcare AI more cautiously, largely because the regulatory question has been unresolved. Broad indexes finished the most recent session mixed: the S&P 500 tracker (NYSEARCA: SPY) closed at $770.19, down 0.39% on the day from a previous close of $773.17, while the Nasdaq 100 tracker (NASDAQ: QQQ) closed at $718.96, up 0.18%. The Dow tracker (NYSEARCA: DIA) ended at $534.08, off 0.53%. Those are the last trades as of 20:00 GMT on Sept. 4, 2026; the market is closed.

None of that is a reaction to TEMPO — no listed company in this story trades on the news. But the index context is the environment in which the next round of digital health financing gets priced, and a credible regulatory on-ramp for generative AI in clinical settings is the sort of structural change that eventually shows up in how those assets are valued.

What to watch next

Three things will determine whether TEMPO becomes a durable pathway or a contained experiment. First, whether the agency expands the cohort beyond four devices and publishes the criteria it used. Second, what conditions attach to participation — the reporting obligations, the monitoring requirements, and what happens to a deployed product if authorization is denied. Third, whether any participant converts pilot participation into full marketing authorization, which is the only outcome that proves the sequence works end to end.

Until then, developers weighing the pathway are making a bet: that reaching patients sooner is worth carrying the regulatory verdict on their back while they do it.

Key facts

• Devices accepted: Four, including AI products from Cadence and Limbic
• Program: FDA TEMPO pilot
• Core provision: Release to patients without marketing authorization
• S&P 500 (SPY): $770.19, -0.39%, last close Sept. 4, 2026, 20:00 GMT

Frequently asked questions

What is the FDA's TEMPO pilot?

TEMPO is a Food and Drug Administration pilot program that allows digital health companies to release certain medical devices to patients before those products have received marketing authorization. Four devices have recently been accepted, including generative AI products from Cadence and Limbic. The pathway is aimed at software that does not fit the traditional freeze-and-review device process.

Which companies have devices in the pilot?

Cadence and Limbic are named as having AI products among the four devices recently accepted into TEMPO. Both are private companies, so there is no publicly traded stock tied directly to their participation. The identities of the other accepted devices were not specified in the reporting on the pilot's initial cohort.

Does TEMPO mean a product is FDA-approved?

No. Participation in TEMPO explicitly means a product can reach patients without marketing authorization. The authorization question remains open while the device is in use. That is the defining feature of the pathway and the reason it represents a shift in how the agency sequences oversight for generative AI software.

Why does generative AI need a different regulatory path?

Traditional device review certifies a fixed product tested against a fixed dataset. Generative models produce variable outputs, can be retrained, and may drift as the user population changes, so a certification of one frozen version may not describe the product months later. TEMPO substitutes supervised real-world deployment and evidence gathering for that upfront certification model.

How does this affect digital health investors?

Marketing authorization has historically been the gate separating spending from revenue for device developers. A pathway allowing earlier patient access shortens the runway needed to reach commercial traction and gives acquirers products with real usage data. The offsetting risk is that a deployed product still faces an authorization verdict, and a denial after deployment is harder to unwind.

What were the major indexes doing around this news?

As of the last trade at 20:00 GMT on Sept. 4, 2026, the S&P 500 tracker SPY closed at $770.19, down 0.39% from a previous close of $773.17. The Nasdaq 100 tracker QQQ closed at $718.96, up 0.18%, and the Dow tracker DIA finished at $534.08, down 0.53%. None of these moves relate to the TEMPO news.

Sources

• STAT+: FDA pilot offers generative AI medical devices a path to patients before they are authorized — STAT News

Photo: Tima Miroshnichenko · Pexels Licence — source