Description

A focused course, tailored for you

The In-House Counsel AI Product Counseling Playbook

What an in-house counsel actually owns when a product team ships an AI feature into a market with no settled rule book.

A product manager just forwarded a launch memo. The feature uses a generative model. Three regulators are named in the risk section. "Counsel sign-off pending" is the only blocking comment. The council vote is Thursday.

$199 one-time

Tailored to your situation. Access within 24 hours. 30-day money-back.

Includes a hand-built implementation playbook delivered alongside course access, generated for your specific situation.

Why this course

In-house counsel inside large consumer-internet employers do not write rules. They translate unsettled rules into one-paragraph answers that engineering, product, and policy can act on this week. The volume of AI-feature reviews has gone from rare to weekly. The DPIA template the privacy team uses was written for ads targeting, not for generative output. The EU AI Act classification question shows up on every launch memo and there is no internal canonical answer. The model-card the research team produces is forty pages long and the product manager wants a yes or no. The retention question for prompts and outputs sits between privacy, security, and product, and counsel is expected to land it. This course gives an in-house counsel the artefacts to walk into a product council with a defensible position on each of those four, in language the engineering lead and the policy lead both accept.

What you walk away with

Scope a DPIA for a generative feature in one sitting using a template that maps to the existing internal privacy review.
Give a defensible EU AI Act provider-versus-deployer classification answer for a shipped feature in a paragraph.
Land a retention position for prompts, outputs, and fine-tuning data that privacy, security, and product all sign.
Hand engineering a model-card review checklist that returns a yes-or-no in an afternoon, not a week.
Write the one-paragraph product-council answer that closes the launch loop without absorbing engineering's risk.

The 12 modules

Module 1. The launch memo, decoded

Open a real-shape AI-feature launch memo and pull out the seven decisions in-house counsel actually owns versus the ones engineering, privacy, and policy own. Learn to read a product manager's risk section as a list of pending counsel questions, not as a finished risk assessment. Practice with a redacted memo and a marked-up version showing exactly where counsel's signature carries weight and where it does not.

Module 2. DPIA scoping for generative features

Walk a DPIA template tuned for generative features through one realistic example end to end. Cover the data categories a generative model touches that an ad-targeting DPIA never asked about, including training data lineage, prompt retention, output reuse, and downstream sharing with vendors. Output is a filled DPIA the privacy team will accept on first read.

Module 3. EU AI Act provider versus deployer

The single question every launch memo now contains. Work through a decision tree that takes a feature description, asks the four classification questions in order, and lands on provider, deployer, or both with the legal reasoning written out. Includes the high-risk Annex III test, the GPAI provider obligation triggers, and the substantial-modification trap that turns a deployer into a provider.

Module 4. Output retention positions you can defend

Prompts, outputs, model fine-tuning artefacts, and chat history each have a different retention story. Work through the four together and produce a defensible internal position that names the retention period, the legal basis, the deletion mechanism, and the audit trail for each. The output is a one-page retention memo the privacy team adds to the internal canonical answers.

Module 5. Model cards as a legal artefact

The research team writes a forty-page model card. Counsel needs the four questions whose answers determine whether the model can ship. Learn the model-card review checklist that turns a long technical document into a structured yes-or-no for fairness, robustness, hallucination rate, and known-failure modes. Includes the language to use when the model card is missing the section that matters.

Module 6. Training data lineage and IP exposure

The training data question gets harder every quarter. Work through the artefacts an in-house counsel needs from the research team to give a defensible answer on copyright, licensing, scraped public data, user-generated content reuse, and the synthetic-data fallback. Output is a training-data legal review template that the research team can fill before requesting counsel sign-off.

Module 7. FTC AI claims and marketing copy review

Marketing wants to call the feature "safe", "trustworthy", or "human-level". Each word invites an FTC follow-up. Walk through the FTC's published guidance on AI claims and learn the substitute language counsel can offer the marketing lead in the same conversation. Includes a one-page marketing copy review checklist for AI-feature launches.

Module 8. The product council answer

Every review ends with a one-paragraph counsel position read aloud at product council. Learn the structure that names what counsel signed, what counsel did not sign, what is conditional on engineering shipping a specific artefact, and what is escalated. The paragraph is the product. Practice writing it for three different feature shapes.

Module 9. Cross-border data and the deployment map

A generative feature deployed globally hits five privacy regimes before lunch. Build a deployment map that shows the legal basis, the cross-border transfer mechanism, and the regulator notification path for each market. Output is the deployment map counsel hands back to the product team as a release-gate artefact.

Module 10. Vendor model and API contract clauses

When the feature uses a vendor model under the hood, the contract is the legal control. Work through the clauses that matter for AI features: training data carve-outs, output IP, indemnity scope, retention and deletion, audit rights, and the change-of-model notification trigger. Output is a vendor AI clause checklist counsel can use in any procurement review.

Module 11. Incident response and regulator letter readiness

A regulator letter arrives about a shipped AI feature. The first forty-eight hours decide what the letter becomes. Walk through the incident-response stack tuned for AI incidents: harmful output reports, model behaviour drift, downstream user complaints, and the policy and PR coordination path. Output is an AI incident response runbook counsel can hand to the SIRT lead.

Module 12. The counsel-side launch playbook

Pull modules one through eleven into one binder counsel uses on every AI-feature review. The playbook includes the decoded launch-memo template, the DPIA, the classification decision tree, the retention memo, the model-card checklist, the marketing review checklist, the product council paragraph structure, the deployment map, the vendor clause checklist, and the incident runbook. The binder is the artefact the next counsel hire reads on day one.

How this addresses your situation

Specific modules that map to what you said you are dealing with.

Module 2, 4, and 9 together answer the privacy review the platform team will send next week.

Module 3 alone answers the EU AI Act classification question on the next launch memo.

Module 5 and 7 together answer the product council on shipped-feature sign-off.

Module 10 and 11 together answer the procurement and incident-response side of a vendor-model deployment.

What you get with this course

Twelve self-paced written modules with worked examples drawn from real launch memos and real DPIA templates.
Downloadable templates for the DPIA, the classification decision tree, the retention memo, the model-card review checklist, the marketing copy review, the deployment map, the vendor clause checklist, and the incident runbook.
The hand-built implementation playbook tailored to your product mix and your reporting line, delivered alongside course access.
Thirty-day refund window if the materials do not match what the role actually faces.

What you will have in hand by Day 1, Week 1, Month 1

Within 24 hours, account provisioned in the Art of Service learning environment and the tailored implementation playbook delivered.

Twelve modules are released together, self-paced, no cohort timeline.

Most counsel finish the binder in three to four working sessions across two weeks.

Before and after

Before

Every launch memo lands the same way. Counsel reads it on Tuesday, sends three follow-up questions on Wednesday, waits for engineering responses on Thursday, lands a position on Friday afternoon, and the product council moves the vote. Each review takes most of a week. The internal canonical answers exist nowhere.

After

Counsel reads the launch memo, runs the DPIA template, runs the classification decision tree, reviews the model card against the checklist, writes the product-council paragraph, and lands the position in one sitting. The canonical answers live in the binder and the next launch reuses them. The product council vote happens on schedule.

What happens if you do not address this

AI-feature launches keep arriving. Each one without a working stack of templates burns a week of counsel time and produces an inconsistent internal position the next launch contradicts. Eventually a regulator letter arrives about a shipped feature and the internal answer trail does not exist. The stack is cheaper to build now than to reconstruct under a regulator's clock.

Who it is for

An in-house counsel inside a large consumer-internet or social-platform employer who reviews AI-feature launches as a recurring weekly load. Title is typically counsel, senior counsel, or associate general counsel within product, privacy, or AI legal. Reads launch memos, attends product councils, signs off on DPIAs, and is on the hook when a regulator opens a letter about a shipped feature. Has a privacy-law foundation, is learning the AI Act and the FTC AI guidance on the job, and wants a working stack of templates rather than another doctrinal article.

Who this is NOT for. Outside counsel writing client memos. Policy staff writing position papers. Engineers building the feature. This is for the in-house lawyer who has to land a product-council answer this week.

How it arrives

Text-based course in the Art of Service learning environment, plus downloadable templates and worked examples for every module, plus the hand-built implementation playbook delivered alongside course access.

Time investment. Eight to twelve hours total across the twelve modules, split however the calendar allows. Each module stands alone and can be done in the gap between two product reviews.

Why $199 is the right number

Doctrinal articles tell you what the AI Act says. CLE webinars tell you what the FTC said last quarter. Outside counsel writes a memo for one feature. This course gives an in-house counsel the working stack of templates the recurring weekly review load actually needs, with the binder the next hire reads on day one.

FAQ

I already know the privacy law. Will this repeat what I know?

The modules assume a privacy-law foundation. The artefacts are the value: DPIA scoping for generative features, the classification decision tree, the retention memo, the model-card checklist, the product-council paragraph structure. The point is to leave with the binder, not the doctrine.

Will the templates match our internal review process?

The templates are designed to plug into an existing internal privacy and product review. The hand-built implementation playbook is tuned to your product mix and reporting line, so the artefacts land in the workflow you already run.

Is this only for the EU AI Act?

No. The classification decision tree covers the EU AI Act because that is the question on every launch memo right now, but the DPIA, retention, model-card, marketing, deployment, vendor, and incident modules apply to AI-feature counseling globally.

30-day money-back guarantee. If after a week of working through the materials this is not what you needed, reply to the receipt email and a full refund is processed. No questions, no forms.

Within 24 hours your account in the learning environment is provisioned and the tailored implementation playbook is delivered alongside it.