ibl.ai Agentic AI Blog

Insights on building and deploying agentic AI systems. Our blog covers AI agent architectures, LLM infrastructure, MCP servers, enterprise deployment strategies, and real-world implementation guides. Whether you are a developer building AI agents, a CTO evaluating agentic platforms, or a technical leader driving AI adoption, you will find practical guidance here.

Topics We Cover

Featured Research and Reports

We analyze key research from leading institutions and labs including Google DeepMind, Anthropic, OpenAI, Meta AI, McKinsey, and the World Economic Forum. Our content includes detailed analysis of reports on AI agents, foundation models, and enterprise AI strategy.

For Technical Leaders

CTOs, engineering leads, and AI architects turn to our blog for guidance on agent orchestration, model evaluation, infrastructure planning, and building production-ready AI systems. We provide frameworks for responsible AI deployment that balance capability with safety and reliability.

Back to Blog

Which AI Platforms Let a Law Firm Preserve Attorney-Client Privilege?

ibl.aiNovember 16, 2026
Premium

A comparison of how Harvey, CoCounsel, ChatGPT Enterprise, Claude for Enterprise and self-hosted ibl.ai handle the facts that decide privilege (where prompts live, training, retention and who holds the logs), read against ABA Formal Opinion 512, Model Rule 1.6 and the Heppner ruling.

The Short Answer

A law firm preserves attorney-client privilege on an AI platform when client information stays confidential in fact: not trained on, not retained by a third party, and reachable only by the people walled in to the matter. Harvey, CoCounsel, ChatGPT Enterprise and Claude for Enterprise promise this by contract. On ibl.ai you own all the code and the data, inside your own perimeter.

That distinction matters because courts have started deciding confidentiality by reading vendor terms. In United States v. Heppner (S.D.N.Y., February 2026), the privacy policy of the tool the defendant used was part of why his AI chats were not privileged.

No platform makes a communication privileged. What a platform can do is keep the facts that privilege depends on, confidentiality above all, under the firm's control.

What does ABA Formal Opinion 512 require before a lawyer puts client information into an AI tool?

ABA Formal Opinion 512, issued July 29, 2024, requires a lawyer to assess the risk that client information entered into a generative AI tool will be disclosed or accessed, under the confidentiality duty in Model Rule 1.6.

For self-learning tools the opinion is direct: "a client's informed consent is required prior to inputting information relating to the representation into such a GAI tool." It adds that "general, boiler-plate provisions to engagement letters" are not sufficient.

Its baseline instruction applies to every tool: lawyers "should read and understand the Terms of Use, privacy policy, and related contractual terms and policies of any GAI tool they use to learn who has access to the information."

The full text is in the ABA Formal Opinion 512 PDF.

Opinion 512 also names a risk inside the firm: a self-learning tool can surface one client's information to lawyers behind an ethical wall. Access control is part of confidentiality, not a separate IT question.

What did the Heppner ruling decide about AI chats and privilege?

United States v. Heppner decided that about thirty-one documents a criminal defendant created with a publicly available AI tool were protected by neither attorney-client privilege nor work product.

Judge Jed Rakoff ruled from the bench on February 10, 2026, with a written memorandum on February 17.

Per Venable's analysis of the ruling, the court found no reasonable expectation of confidentiality because the tool's consumer privacy terms allowed inputs to be retained, used for training and disclosed to third parties.

The defendant used the tool on his own, without counsel's direction, which also defeated work product. The court did not rule on enterprise tools or attorney-directed use.

The lesson for a firm choosing a platform is narrow and practical: when confidentiality is tested, someone will read the terms that govern the tool, and those terms will be quoted back.

How do Harvey, CoCounsel, ChatGPT Enterprise and Claude for Enterprise handle client data?

Harvey, CoCounsel, ChatGPT Enterprise and Claude for Enterprise each publish no-training commitments and some form of retention control, according to their own pages read on November 16, 2026.

They differ on who sets retention, and all four hold your data on infrastructure you do not run.

Platform Training on client data Retention Where prompts and documents live Nature of the protection
Harvey "Never used to train AI models unless explicitly authorized by both the customer and Counsel AI" Zero retention at subprocessors: "only processed ephemerally" Harvey's environment on AWS and GCP in the US, EU or Australia Contract and vendor policy
CoCounsel (Thomson Reuters) Not used to train CoCounsel; partners "contractually prohibited" from training Zero-retention calls to OpenAI and Google; retention "customized to fit customer needs" Thomson Reuters' environment Contract and vendor policy
ChatGPT Enterprise Not used for training by default Workspace owners set custom retention (OpenAI's admin guidance gives a 90-day minimum) OpenAI's environment Contract and vendor policy
Claude for Enterprise Anthropic says it has "never trained on enterprise data without explicit permission" Indefinite unless an owner sets a custom period, minimum 30 days Anthropic's environment Contract and vendor policy
ibl.ai (self-hosted) Nothing leaves for training; the stack is yours Your database, your retention policy, your deletion Your own cloud, on-premise or air-gapped network Infrastructure you control, plus the source code

Sources: Harvey subprocessor FAQs, CoCounsel Essentials, OpenAI business data, OpenAI Academy on data governance, Claude Enterprise retention controls, Anthropic Enterprise Frontier Safeguards.

These are real, serious commitments, and for many matters they are enough. The column that differs is the last one.

Why does a contractual privacy promise differ from infrastructure a law firm controls?

A contractual privacy promise differs from infrastructure a law firm controls because the vendor can change its terms, a court can order the vendor to preserve data, and the firm learns about either after the fact.

Infrastructure the firm runs changes only when the firm changes it.

The terms do move. Anthropic introduced 30-day data retention starting with Fable 5, then announced on September 1, 2026 that eligible enterprise customers would receive zero retention on Fable 5 and Fable 5.1 until its new Enterprise Frontier Safeguards are ready.

Courts can override retention settings too.

In the New York Times litigation, OpenAI was ordered in May 2025 to preserve consumer and API output logs it would otherwise have deleted (Enterprise was excluded). The order was largely lifted in October 2025, with already-preserved logs still held.

We covered what that meant for firms in NYT v. OpenAI discovery and the law firm AI checklist.

Opinion 512 asks the lawyer to know "who has access to the information." With a hosted vendor the answer is a document that can be revised. With a self-hosted platform it is your own access list and your own logs.

Does a self-hosted AI platform guarantee attorney-client privilege?

A self-hosted AI platform does not guarantee attorney-client privilege. Privilege depends on the communication, the relationship and counsel's direction, as Heppner shows. What self-hosting removes is a third party from the confidentiality analysis.

One honest limit: a self-hosted platform that calls a hosted model still sends each prompt to that model provider, under that provider's terms.

For the most sensitive matters, run an open-weight model inside the same perimeter, or on a fully air-gapped network; see air-gapped AI for law firms.

ibl.ai is the agentic AI platform where you own all the code and the data. Firms run it in their own cloud, on-premise or air-gapped, with role-based access so a matter's agents and documents stay behind its ethical wall, and audit logs the firm holds.

Because ibl.ai is model-agnostic, the same firm can use a hosted frontier model for public research and a local model for privileged drafting. For the Rule 1.6 detail, see ABA Model Rule 1.6 compliant AI, and for pricing against Harvey, the Harvey AI alternative.

ibl.ai is family-owned and operated from New York, NY, and our engineers deploy the platform alongside a firm's own IT team.

Why does owning the AI stack matter?

ibl.ai is the agentic AI platform where you own all the code and the data. You self-host the entire stack inside your own perimeter, run it model-agnostic across any LLM and switch anytime, and pay by usage with no per-seat pricing β€” so you can deploy anywhere: your cloud, on-premise, GovCloud, or fully air-gapped.

  • You own all the code and the data

    Full source code under a perpetual license, running on your infrastructure. Not API access to someone else's platform β€” the stack itself is yours.

  • Model-agnostic

    Run any LLM β€” Claude, GPT, Gemini, Llama, Command, or your own fine-tune β€” and switch providers without rewriting the platform.

  • No per-seat pricing

    Usage-based billing against a budget cap you set. Cost tracks what your organization actually uses, not how many people you employ.

  • Deploy anywhere

    Your cloud, your VPC, on-premise, GovCloud, or a fully air-gapped network with no outbound connectivity.

1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University.

ibl.ai is family-owned and operated from New York, NY β€” a U.S.-headquartered, domestically-owned long-term partner, not a vendor that sells licenses and moves on.

See the ibl.ai AI Operating System in Action

Discover how leading universities and organizations are transforming education with the ibl.ai AI Operating System. Explore real-world implementations from Harvard, MIT, Stanford, and users from 400+ institutions worldwide.

View Case Studies
Work with our team

Pilots, deployment, and full ownership

Every engagement is scoped to the work: you integrate ibl.ai with your own data, deploy it on your own infrastructure, and the engineering hours scale with the scope β€” so the price tracks the scope, not your headcount. Most customers keep us on afterwards for platform updates, new agents and integrations, and ongoing services.

Start here

Pilot

from $15K

fixed scope Β· fixed timeline

A time-boxed proof of value on your real data β€” not a slide deck.

Best for: Teams that want to see ibl.ai working before committing.

  • Deployed on your infrastructure or our cloud
  • 1–2 production agents wired to a slice of your data
  • One integration (LMS / SIS / SSO / data source)
  • Weekly working sessions with our engineers
  • Pilot fee credits toward a full engagement
Scope a pilot
Most common

Integration & Deployment

$25K – $80K

one-time fee Β· scoped to the work

Full deployment integrated with your data and systems. Engineering hours scale with scope.

Best for: Organizations rolling ibl.ai out across a department, campus, or business unit.

  • Platform deployed in your VPC, on-prem, or air-gapped
  • Integrated with your data + identity (SSO / SAML)
  • Multiple custom agents built to your workflows
  • Engineering hours proportional to scope
  • You own the data Β· run any LLM you choose
Plan a deployment
Full ownership

Codebase Transfer + Custom AI Engineering

Custom quote

perpetual license Β· you own the stack

We transfer the full source code. You own and self-host the entire platform β€” outright.

Best for: Organizations and enterprises that benefit from perpetual ownership and sovereignty.

  • Complete source-code transfer + perpetual license
  • Dedicated AI engineering team on your roadmap
  • Custom agents, models, and integrations to spec
  • Air-gapped capable Β· zero vendor lock-in
  • Family-owned, New York–based long-term partner
Talk about ownership
You own the code and data Run any LLM β€” Claude, GPT, Gemini, Llama Family-owned & operated from New York, NY