Freed

Ambient capture that turns a visit into a structured SOAP note is the product itself, and every note comes from the model. The three products built around the scribe are model driven too. Clinical Evidence answers clinical questions with cited answers from more than 50 sources, the Coding Assistant suggests ICD-10 and CPT codes at the highest justified E/M level inside the scribe workflow, and Front Desk is an AI receptionist that answers patient calls during lunch, after hours or at overflow. The scribe also learns from edits, so a clinician can correct a note and ask Freed to carry that style into future notes.

The clinician keeps complete control over what goes into the note and how it is finalized. The final note can be edited or deleted at any time, and a quality check runs before the audio recording is deleted, which implies a verification stage between generation and finalization.

Edits also teach the system: a clinician can correct a note and ask Freed to learn that style for next time, so the review step doubles as a training signal.

Low confidence passages are not flagged to the clinician reviewing the draft, so review is exposed to the automation bias that fluent, usually correct drafts produce.

Freed describes its models by what they learn from and what they recognize. Its AI is trained only on de identified notes, and protected health information is never used for training. Speech recognition is described as trained for clinical use and deriving medications, dosages and abbreviations, and Clinical Evidence draws on more than 50 sources with linked citations. The homepage claims 98 percent recall on medical terms across more than 30 specialties, tested against thousands of clinical concepts.

That recall figure is the only accuracy number published, and it comes without a test set, method or error breakdown. No model class, version, foundation model provider, architecture or update policy is named, and nothing separates what is built in house from what is licensed. Microsoft Azure is named as infrastructure, which identifies the cloud but not the model, and the 50 sources are counted rather than listed.

Microsoft Azure hosts the product, with all data processed and stored in the United States under a business associate agreement with Microsoft, so both the cloud and the legal footing for content sent to it are named. Subprocessor governance is set as a rule: every vendor that may process patient information must be HIPAA compliant and sign an agreement with Freed.

The foundation model behind the notes, and the subprocessors that have met that rule, are not named.

A recall figure, adoption counts and one detailed customer account make up the case for Freed's results. The homepage claims 98 percent recall on medical terms across more than 30 specialties, tested against thousands of clinical concepts, and reports more than 26,000 clinicians, more than 1,300 clinics and 32,589,627 patient visits transcribed in 2025, alongside quotes from individual clinicians.

The Camarena Health case study adds a time result. The community health center runs 24 clinical sites in California's Central Valley, and more than 50 of its clinicians now use Freed. In 2025 they documented 96,067 visits with it, and documentation fell from roughly 10 minutes a visit to about 2 minutes of review, which Freed works out to approximately 12,800 hours returned in a year.

Freed encourages prospective buyers to run their own pilot through the free trial and generate their own data. No peer reviewed study, controlled evaluation or research page is available.

Recordings are kept only as long as the note needs them. Audio is held until the note and its quality checks are complete, then deleted automatically. Clinicians can edit or delete notes at any time, or set them to delete automatically after 30 days. Everything stays in the United States, on Microsoft Azure under a business associate agreement with Microsoft. Every vendor that may process patient information must be HIPAA compliant and sign an agreement with Freed, and cryptographic modules follow FIPS PUB 140-2.

The training commitment is conditional. Freed states that protected health information is never used for AI training and that its AI is trained only on de identified notes. De identified notes still come from customer encounters, so customer content does shape the model, with de identification as the safeguard rather than an exclusion. The control is an opt out rather than an opt in default, which leaves the customer to exercise it. Never training on protected health information is a narrower promise than never training on customer data.

Protected health information is governed by a business associate agreement that anyone can read. It is Schedule A to the Platform Terms of Use, last updated August 13, 2026, names Freed, Inc. as Business Associate, and sits in the help center, so a clinician can read it before creating an account.

The agreement commits Freed to a written privacy and security program with administrative, technical and physical safeguards, breach notification within ten days of discovery, a duty to report and mitigate improper uses, minimum necessary limits, access for the Secretary to internal practices, and return or destruction of protected health information on termination, including what subcontractors hold. Section 2(e) binds any subcontractor that receives protected health information to the same restrictions and conditions that apply to Freed.

Coverage attaches on use rather than on plan. Section 7 of the Terms says that by providing Your Data the customer agrees to be bound by the agreement, so even the $39 Starter plan carries it. The help center adds that a separately signed copy can be requested, including during the trial, and the Groups plan carries an organization wide agreement. Freed's marketing says it maintains business associate agreements with all enterprise customers, which understates what the terms give a solo clinician.

SOC 2 Type 1 and Type 2 certification, described as independent third party audits, sits alongside HIPAA and HITECH compliance, and the controls behind them are named. Cryptographic modules follow FIPS PUB 140-2, a United States federal standard. Data is encrypted with TLS 1.2 to 1.3 in transit and AES-256 at rest, hosted on Microsoft Azure. OWASP secure coding standards are enforced with regular audits, and Azure Security Center and Drata handle continuous monitoring and vulnerability scanning.

Operational controls are described as well. Infrastructure is managed as code with every change reviewed before deployment, every engineer takes secure development training, all staff pass background checks and annual HIPAA and privacy training, and access uses single sign on and multifactor authentication.

Vendors are governed by rule. Freed holds a HIPAA compliant business associate agreement with Microsoft, requires every vendor that may process patient information to be HIPAA compliant and sign an agreement with Freed, and reviews vendor security practices on an ongoing basis.

Freed holds no FDA clearance, and its terms address regulatory status for the expanded product rather than the scribe alone. The Platform Terms of Use, last updated August 13, 2026, say the platform, the content and the output are not intended to diagnose, treat, cure or prevent any disease or health condition, and are not a substitute for professional medical advice, diagnosis or treatment.

Section 11 names the clinical decision support feature and governs it separately. Freed's Clinical Evidence is defined there as a tool that analyzes customer data, applies evidence based guidelines and generates recommendations to assist in diagnosis, treatment planning and other clinical decisions. The feature is intended solely to support and not direct clinical care, lets the user review the basis of its recommendations, and must only be used as an adjunct to clinician reasoning. Relying on it without human oversight to create, finalize or submit clinical documentation, billing or coding is prohibited.

Two surfaces sit closer to the edge of that position. Codes are marketed as suggested at the highest justified evaluation and management level, and Front Desk, the AI receptionist that answers patient calls, puts the system in direct contact with patients rather than clinicians under a separate Platform Terms of Use document.

Community clinics are Freed's market, and its Camarena Health case study describes notes that adapt to multiple languages, even within the same visit. Speech recognition accuracy varies with accent, dialect, speech rate and vocal characteristics, and the patients most affected overlap with those already underserved.

The one published performance figure, 98 percent recall on medical terms across more than 30 specialties, is an aggregate with no breakdown by speaker, accent or language. Security engineering is described standard by standard, from FIPS PUB 140-2 cryptography to OWASP secure coding and vulnerability scanning with Drata, and model behavior gets no comparable treatment.

The product is marketed with specialty awareness, including an understanding of sensitive conversations in fields such as behavioral health. That is a content handling claim rather than a performance disclosure. No responsible AI statement, bias or fairness position, subgroup or demographic analysis, error taxonomy, model monitoring description or evaluation output is published.

Clinicians can edit and delete notes at any time, and Section 11 of the terms lets the user review the basis of each Clinical Evidence recommendation, so the person who signs the note can fix what the system got wrong and see what a recommendation rested on.

Liability is narrow. Section 13 caps direct damages at the greater of $100 or three months of fees, so a clinician on the entry plan is capped at little more than $100. The platform is provided as is, with accuracy, completeness and currency expressly disclaimed, and Freed states it will have no liability for any harm arising from use of the platform or the output. Section 16 runs indemnification one way, from the customer to Freed. During a trial, the main way in, the terms carry no warranty, indemnity, support or other obligation at all, and Section 6 of the business associate agreement confirms that no one other than the parties is a third party beneficiary, so the recorded patient has no route.

Speech recognition is described as clinically trained to derive medications, dosages and abbreviations, a capability the terms do not warrant, and medication and dosage extraction is where an error carries the most weight.

EHR Push is a Chrome extension that maps and transfers a finished note into a browser based EHR with one click. Freed says it works with dozens of browser based systems and works best on SimplePractice, PracticeFusion, Tebra, Elation, Therapy Notes, Athena, Optimantra, Kipu, Aarista and DrChrono, with support for new EHRs added monthly. Installing it takes less than a minute and needs no IT support, and it is part of the Premier plan.

The transfer is note text. There is no native EHR API connection and no structured, field level write back, so discrete data such as problems, orders or codes does not populate the chart.

Freed runs in a web browser or mobile app with no additional hardware, supports offline and low connectivity use, and needs no IT implementation, so a clinician can deploy it from a free trial without an institutional purchase. All data is processed and stored in the United States, in Microsoft Azure data centers, under a HIPAA compliant business associate agreement with Microsoft, and Azure's high availability infrastructure is cited for continuity.

Support is live, with scheduled calls, office hours and recorded walkthroughs, plus priority support and a dedicated account manager on the Groups plan. Residency is stated at country level, with no region or failover location given, and no uptime commitment is published.

Self serve pricing is published in full, so a clinician can see the total cost, compare plans and start without talking to sales. Starter costs $39 a month for one clinician and up to 40 notes. Core costs $79 a month with unlimited notes. Premier costs $119 a month, or $104 a month billed annually, and adds EHR Push, visit summaries, patient instructions and letters, and ICD-10, CPT and E/M code suggestions. Front Desk, the AI receptionist, starts at $149 a month.

A seven day free trial opens the Premier features with no credit card needed. Eligible students, residents, trainees and fellows can get 12 months of Core at no cost, or up to 50 percent off monthly Premier for 24 months. The Groups plan starts at two clinicians and is priced by headcount. Smaller groups may pay the standard individual rate, and larger organizations may qualify for custom pricing.

The terms describe the same commercial shape. Subscriptions run monthly or annually, purchase is self serve through a third party payment processor, cancellation happens from the account billing page, and prepaid fees are not refunded.

Independent and community practices are the target, with primary care and general practice at the center. Freed's site has specialty pages for nurse practitioners, family medicine, internal medicine, psychiatry, mental health, pediatrics, functional medicine and OB/GYN. Every plan includes specialty specific templates, and the homepage claims 98 percent recall on medical terms across more than 30 specialties. At Camarena Health, clinicians across more than 10 specialties use it, with notes that adapt to multiple languages, even within one visit.

The company also names the setting it leaves out, stating that it is not made for massive healthcare systems. No results by specialty are published, and procedure heavy and inpatient documentation are not covered by a named deployment.