Recording Consent Without a Consent Conversation

TL;DR

The law does not require you to open a call by asking permission. In the strictest state in the country the requirement is that the other party knows, before capture starts — a standard you satisfy with notice, not with a negotiation.
Two rules set the shape: consent has to be prior (Javier v. Assurance IQ), and in Massachusetts the offence is secret recording rather than unconsented recording (Commonwealth v. Hyde). Put the fact in front of people early and the call itself needs no ceremony.
Build it in layers — scheduler, calendar invite, first email, contract — so no single missed touchpoint leaves someone uninformed. Each layer is about ten minutes of work and you do it once.
Say it once on the call, in one sentence, only when someone plainly has not seen the notice. That is not a consent conversation, and it takes four seconds.
In-person capture is the exception. There is no invite layer for a coffee meeting, so that one genuinely does need a spoken heads-up.

Every team that takes this seriously arrives at the same unhappy place: a script at the top of every call that begins "before we get started, I just need to let you know…". It kills the opening, it makes a routine meeting feel like an interrogation, and reps quietly stop doing it within about three weeks.

The good news is that the ceremony is not what the law asked for. It is what people assume the law asked for. The actual requirement, across the statutes that matter, is narrower and much easier to design around.

This is operational guidance, not legal advice. Which states your buyers sit in changes the analysis — take your specific setup to counsel.

What the requirement actually is

Two decisions define the shape of the problem, and between them they describe a process most teams would happily run. We cover both in detail in the lawsuits that wrote the rules.

Consent has to come first. In Javier v. Assurance IQ, the Ninth Circuit held that California's § 631(a) requires the prior consent of all parties and that retroactive consent does not fix a recording that already happened. Agreeing to a privacy policy after the capture began was worth nothing.

Read that as a constraint and it kills the easy options — the privacy policy nobody opened, the line in the recap email. Read it as a specification and it tells you exactly what to build: the notice has to be earlier. It says nothing at all about the notice being spoken, or happening on the call, or taking the form of a request.

And in the strictest state, the offence is secrecy. Massachusetts has the toughest recording statute in the United States — Commonwealth v. Hyde affirmed a criminal conviction for secretly recording a traffic stop, with no exception even though the encounter was public. But read what § 99 actually prohibits, as quoted in that opinion:

"The term 'interception' means to secretly hear, secretly record… by any person other than a person given prior authority by all parties to such communication." — G. L. c. 272, § 99 B 4

Secrecy is an element of the offence. The carve-out is prior authority from all parties. And the court's holding tracks both: § 99 bars secret recordings made "without their permission or knowledge."

Note that Massachusetts, entirely independently of the Ninth Circuit, also says prior. Two of the strictest authorities in the country converge on the same instruction: get there first.

That is a documentation problem, and documentation problems are solvable once rather than performed daily.

The layered model

Do not look for the single touchpoint that discharges the obligation. Build overlapping layers so that a buyer would have to miss all of them to be surprised — which is both better practice and a far better story if anyone ever asks you to reconstruct what a participant knew.

Layer 1 — the scheduler (the highest-leverage ten minutes you will spend)

If meetings are booked through Calendly, Chili Piper, HubSpot or similar, put the line on the booking page and in the confirmation email. This layer is the best one you have, because the person has to pass through it to get the meeting, and it lands well before the call.

Meetings are captured by an AI assistant so we can share accurate notes afterwards. Let us know any time if you'd prefer we didn't.

Layer 2 — the calendar invite

Put it in the invite description, near the top where it survives truncation in the notification preview. Every attendee sees the invite; most see it more than once.

Note: I use an AI notetaker on my calls. Just tell me if you'd rather I turn it off.

The limitation is worth knowing: a forwarded invite, someone dialling in from a colleague's link, or a late addition may never have seen the description. The invite is your default layer, not your only one.

Layer 3 — the first email in the thread

One line in the email that confirms the meeting. Not the signature block — signature text is visual noise that people genuinely do not read, and "it was in my signature" is a weak account of what someone knew.

Layer 4 — the contract, for existing customers

For accounts you already have a paper relationship with, the cleanest place is the MSA or an addendum: a clause stating that calls may be captured for note-taking, naming the vendor, and describing retention. Once signed, it covers the recurring meetings with that account and you stop thinking about it.

This does not extend to the customer's guests who are not party to the agreement. A QBR where the customer brings a new executive is a call where the invite layer is doing the work.

Layer 5 — the four-second version, on the call, when it is needed

This is what replaces the script. Not at the top of every call. Only when the person plainly has not seen anything — a cold inbound, someone who just joined from a forwarded link, a new face on the customer's side.

I've got my AI notetaker running so I'm not typing while you talk — shout if you'd rather I didn't.

That is one sentence, it is said while screens are loading, and it does not sound like a disclaimer. The critical property is that it is factual and casual rather than permission-seeking. "Is it OK if I record this?" invites a negotiation and makes the tool sound like an imposition. Stating it plainly, with an easy exit, almost never draws an objection — and when it does, you wanted to know.

Wording that does not spook people

The single biggest driver of awkwardness is not the disclosure. It is legal register. Compare:

Don't	Do
"By joining this meeting you consent to the recording and processing of your personal data."	"I use an AI assistant to take notes on these calls."
"This call is being recorded for quality and training purposes."	"So I'm not typing the whole time, I've got a notetaker running."
"Do I have your permission to record this call?"	"Shout if you'd rather I turn the notes off."

Three principles behind the right-hand column:

Lead with why, not with the legal category. "So I can pay attention instead of typing" is a reason people recognise and approve of. "For compliance purposes" sounds like something is being taken from them.
Offer the exit unprompted. Counter-intuitively this reduces objections rather than inviting them, because it signals you are not trying to get away with something. It also does real legal work: an easy, obvious objection path strengthens any argument about what a reasonable participant understood and accepted.
Do not say "recording" if what you mean is "notes", and do not say "notes" if it is in fact recording audio. Both directions cause problems — the first makes people picture a tape they will never see, the second is a misdescription that will read very badly if the question ever becomes adversarial. Describe what actually happens.

What does not work

Each of these fails for a specific, citable reason.

The privacy policy on your website. Fails Javier if the buyer encounters it after capture, and fails on the facts regardless — a policy the person never opened is not knowledge.

A disclosure in the post-call recap. The clearest possible case of retroactive consent.

"We're bot-free, so there's nothing to disclose." Whether a participant tile appears in the meeting is not an element of any of these statutes. They ask whether a communication was captured. If your tool captures audio, the absence of a bot changes how visible that is — and visibility is precisely what these cases ask you to increase. More on why this reasoning fails.

"Our reps are in a one-party state." Kearney v. Salomon Smith Barney held California's statute reaches recordings made lawfully in Georgia, of calls into California. Your buyer's location is the one that matters.

"I'm on the call, so it's my conversation to record." Smith v. LoanMe rejected exactly this for California's § 632.7. An eighteen-second call was enough.

A beep. LoanMe had one. It did not save them.

Where this does not apply: in the room

The companion-app case — a phone or watch capturing an in-person conversation — is the one mode where the layered model mostly collapses. There is no calendar invite for the coffee before the QBR, no booking page for the corridor conversation, and often no meeting at all in the formal sense. The people present have had no opportunity to learn anything in advance.

So in-person capture genuinely does need the spoken version, and it should be said before you start rather than after:

Do you mind if I record this so I get the details right? Happy not to.

Two things make that materially different from the video-call script. It happens rarely, because most of your meetings are not in person. And it is a normal thing to say to someone sitting across a table, in a way that a formal disclosure at the top of a Zoom is not.

Also worth being blunt: in-person is where all-party statutes are at their most dangerous, because a conversation over dinner carries a far stronger expectation of privacy than a scheduled business call. Flanagan asks whether a party reasonably believed nobody was recording. At a restaurant, they certainly did. If you are not prepared to say the sentence out loud, do not capture.

When someone objects

Rarely, and it is a good outcome when it happens, because the alternative was capturing someone who did not want to be captured.

Turn it off, say so, and confirm it is off. Do not argue about whether the tool technically counts as recording, do not explain the architecture, and do not promise anything about past calls you have not verified. We wrote a fuller playbook for this: a customer says no recording — what are your options.

One thing worth building in advance: know how to produce meeting notes without capture, so "no" costs you the automation rather than the meeting.

A checklist you can finish this afternoon

Add the line to your scheduling tool's booking page and confirmation email. (10 min, covers most inbound)
Add it to the calendar invite template your team uses. (10 min)
Add it to the meeting-confirmation email template. (5 min)
Draft the one-sentence on-call version and put it in onboarding. Do not make it a mandatory script. (15 min)
Add a recording clause to the MSA template for new customers, and to the renewal addendum for existing ones. (counsel, 30 min)
Write down the procedure for turning capture off mid-call, and make sure every rep has done it once. (15 min)
Decide your in-person rule and make it explicit: ask out loud, every time, or do not capture. (5 min)
Ask your vendor, in writing, whether meeting content is used for anything beyond returning your own notes — model training included. Put the answer in the contract. (This is the question the Otter litigation turns on.)

Items 1 to 3 are the ones that remove the awkwardness, because once they are live, item 5 is needed only occasionally and item 4 becomes a casual aside rather than a recital.

Frequently asked questions

Do I need to get consent from every attendee, or just the organiser? All-party statutes mean all parties. That is the reason the invite layer matters — it reaches everyone on the invite, not only the person you negotiated with. It is also why an unexpected attendee is the situation calling for the four-second version.

If notice is enough in Massachusetts, is notice enough everywhere? No, and this is the most important caveat here. Massachusetts § 99 is written around secrecy, which is why knowledge satisfies it. Other all-party states are written around consent, where notice plus continued participation is a strong position but not a statutory safe harbour. The practical posture is the same either way — clear prior notice plus an easy way to object — but do not tell yourself Hyde is a nationwide permission slip.

Does silence after a disclosure count as consent? Implied consent from continued participation after clear notice is a recognised argument, and it is stronger the more clearly and earlier the notice was given. It is an argument, not a guarantee, and it is weakest exactly where you would expect: someone who never saw the notice.

Is it enough that our platform shows a recording indicator? It helps, and it is genuinely good evidence of knowledge under a Hyde-style standard. It is also the layer you control least — indicators differ across Zoom, Teams and Meet, and may not appear at all for capture that happens outside the platform. Treat a platform indicator as a bonus layer, never the primary one.