Your business is introducing AI to help people work faster, find information more easily and make better decisions. But as AI gains access to corporate information, how confident are you that existing legal obligations and information controls are being maintained?
AI is increasingly becoming part of everyday business operations. Employees are using it to summarise documents, analyse contracts, search internal repositories, prepare reports and support decisions. As organisations move beyond individual tools towards AI systems connected to corporate information, the opportunities are significant.
But so are the questions that in-house legal teams need to be asking.
What happens when an AI assistant can retrieve confidential information from multiple departments? Could legally privileged material become available to people who shouldn't see it? Are retention and preservation requirements still being respected? And can you demonstrate that the information being accessed and used is subject to appropriate controls?
These aren't simply technical questions. They have potential legal, regulatory and contractual consequences, even when the technology is selected, deployed and managed by another department.
Legal doesn't need to own the AI strategy, but it does need to help ensure that the information AI uses is governed in a way that protects the business and meets its obligations.
AI is creating new challenges for familiar legal responsibilities.
For in-house legal teams, the risks associated with AI are not entirely new. Confidentiality, legal privilege, data protection, retention, preservation and accountability have always been important considerations when managing corporate information.
What is changing is how that information can be accessed, combined, processed and used.
An employee might previously have needed to locate a document, obtain permission to access it and review its contents manually. An AI assistant connected to corporate repositories may now be able to search across multiple information sources, retrieve relevant material and produce a summary within seconds.
That can improve efficiency, but it also means weaknesses in existing information management practices may have more significant consequences.
The Association of Corporate Counsel (ACC) addressed these challenges in its May 2026 programme on AI in legal practice. It identified privilege, discovery obligations, professional responsibility and the use of generative and agentic AI as important areas of consideration for in-house counsel.
The underlying message is clear: the introduction of new technology doesn't remove existing legal obligations. It changes the circumstances in which those obligations must be managed.
For Legal, this means becoming involved before AI systems are given access to corporate information, rather than waiting until questions arise about how that information has been used.
Five information governance risks Legal need to consider.
-
Corporate information carries different levels of sensitivity and different restrictions on access and use. Commercial agreements, legal advice, employee records, investigation materials and confidential business communications may all require protection.
The challenge is ensuring those protections remain effective when information is accessed through AI.
Consider a corporate AI assistant connected to document repositories used by Legal, HR, Finance and other departments. If permissions are too broad, outdated or inconsistently applied, users may be able to retrieve information they were never intended to access. Even where the AI tool is functioning as designed, the underlying access arrangements may create a problem.
For in-house legal teams, legal professional privilege adds another consideration. Information that forms part of privileged communications or advice may require particular care when it is uploaded, processed or made accessible through AI services.
The UK's Solicitors Regulation Authority addressed these concerns in its August 2026 warning notice on the misuse of AI. The SRA highlights risks to confidentiality and privilege, including the importance of appropriate contractual and technical safeguards when sensitive information is processed by AI tools. It also specifically cautions in-house solicitors that AI systems developed for general business use may not have been designed for legal work.
Although the SRA's requirements apply to the solicitors and firms it regulates, the underlying concern is relevant to corporate legal teams across jurisdictions.
Privilege isn't automatically lost simply because AI is involved, nor is it automatically protected because an AI system has been approved by IT. The position depends on the applicable law, the nature of the communication, the system being used and how information is disclosed or processed.
Legal needs to work with IT, Information Security, Privacy and business owners to establish which information can be made available to AI, under what conditions and with which restrictions.
Approving an AI tool for business use should not automatically mean approving every category of corporate information for use within it.
-
One of the less obvious risks of AI is that it can make historic information easier to find and use, including records that should already have been reviewed for retention or disposal.
Corporate repositories may contain years of correspondence, superseded contracts, employee records, investigation materials and information retained without a clear continuing purpose. Some of it must be preserved for legal, regulatory or contractual reasons, but other information may no longer need to be kept.
If these repositories are connected to AI tools, information that was previously difficult to locate can become readily available for searching and analysis.
This creates a challenge for Legal. The fact that historic information might be useful to an AI system doesn't necessarily provide a legitimate reason to retain it indefinitely.
The Information Commissioner's Office guidance on AI and data protection reinforces this point. Under UK data protection law, the principles of data minimisation and storage limitation continue to apply when personal information is processed using AI. Personal data should be limited to what is necessary for the relevant purpose and should not be kept longer than required, subject to applicable exceptions.
For in-house teams, the question is how existing retention requirements are being applied before information becomes available to AI.
Can you identify records that should be retained? Are preservation obligations taken into account? Do you have processes for reviewing and approving disposal? And when information is disposed of, are relevant AI indexes, cached content and downstream copies also addressed?
Defensible disposition becomes particularly important. The aim is not simply to delete unnecessary information, but to demonstrate that disposal decisions were made in accordance with applicable obligations, agreed policies and appropriate authorisation.
If information shouldn't still be held, making it available to AI doesn't resolve the problem. It can make the need for effective governance more urgent.
-
When litigation, an investigation or another preservation obligation arises, Legal needs confidence that relevant information can be identified and protected.
AI introduces additional considerations because information may be copied, indexed, summarised or processed outside its original repository. Depending on the technology and configuration, new records may also be created through prompts, outputs, logs and AI-generated documents.
This raises questions about how preservation requirements apply to information moving through AI-enabled workflows.
For example, if information is subject to a legal hold, how is that restriction reflected in connected systems? Could an automated retention process dispose of relevant material? Are AI-generated records or logs potentially relevant to an investigation, and do existing procedures address them where preservation is required?
Not every AI interaction or generated record will necessarily be subject to preservation obligations. The answer depends on the circumstances and applicable legal requirements.
However, Legal needs to understand the systems and processes involved well enough to identify when additional preservation measures may be necessary.
The ACC's 2026 programme on AI in legal practice highlights discovery, evidence authentication and professional responsibility as emerging areas requiring attention. The practical implication is that in-house teams should review whether established litigation readiness and legal hold procedures remain adequate as AI becomes part of business operations.
This is where information governance plays an important role. Clear ownership, consistent retention processes, appropriate legal hold controls and documented workflows make it easier to manage preservation requirements across systems.
The objective isn't to preserve everything AI touches. It's to ensure that information subject to legal obligations is identified and managed appropriately, regardless of the technology used to access or process it.
-
Introducing AI into business operations can change how information is processed, particularly when tools are provided or hosted by third parties.
An AI application may receive documents, extract relevant information, generate new content or transmit data to services outside your organisation's direct control. Depending on the provider and configuration, prompts, outputs and other information may also be stored or retained.
For Legal and Privacy teams, these arrangements raise familiar questions about lawful processing, confidentiality, contractual obligations, data transfers, retention and accountability.
But the answers may be less straightforward when AI services are involved.
The ICO's guidance on security and data minimisation in AI emphasises that businesses need to understand the personal information being processed, assess whether it is necessary and consider appropriate safeguards, including when using third-party AI providers.
The UK's National Cyber Security Centre also recommends that organisations establish controls over the information AI systems can access, maintain appropriate security arrangements and consider risks throughout the AI lifecycle.
For in-house legal teams, this means becoming involved in relevant supplier assessments, contractual reviews and information processing decisions.
Can the provider explain how information is used, stored and protected? What restrictions apply to further processing or model training? Are retention and deletion arrangements appropriate? Do existing contracts and information sharing agreements permit the intended use?
You don't need to perform every technical assessment yourself, but you should be confident that the relevant legal and contractual requirements have been considered and that responsibility for implementing the necessary safeguards is clear.
-
As AI becomes more widely used, questions about accountability become increasingly important.
If confidential information is exposed, an inappropriate document is used or an AI-generated output contributes to a problematic decision, can you establish how the information was accessed, which controls applied and who was responsible?
This is not simply a question of monitoring the AI system. It also depends on how the source information was governed.
You may have an approved AI platform, documented acceptable-use policies and technical security controls. But if information ownership is unclear, retention requirements aren't consistently applied or access permissions are poorly managed, important gaps remain.
Effective information governance helps address these weaknesses by establishing ownership, applying policies and maintaining evidence of information management decisions.
At the same time, AI systems require their own controls covering logging, monitoring, output verification and human oversight. Information governance supports these activities but cannot replace them.
For Legal, the priority is ensuring that information-related obligations and responsibilities are understood across both frameworks.
You need to be able to demonstrate not just that AI was approved for use, but that appropriate decisions were made about the information it was allowed to access.
Where should Legal sit in the AI governance conversation?
One of the challenges for in-house legal teams is determining how involved they need to be in an AI programme that may be led by IT, Innovation, Digital Transformation or another business function.
Legal doesn't need to own the AI strategy or take responsibility for every technical decision. But waiting until an AI system has been selected and deployed can leave important information governance questions unresolved.
The most effective approach is to establish Legal's role as part of a wider governance framework, working alongside IT, Information Security, Privacy, Risk, Compliance, Information Governance and business owners.
Responsibilities will vary, but there are some clear distinctions.
IT and Security should assess technical controls, integrations and system access. Business owners need to establish the intended use and value of information. Privacy and Compliance teams should consider applicable regulatory requirements, while Information Governance and Records Management help establish retention, disposition and lifecycle controls.
Legal's role is to identify the legal and contractual obligations associated with the information, advise on confidentiality and privilege, establish relevant preservation requirements and help ensure appropriate accountability.
These responsibilities should be coordinated, particularly where decisions involve multiple systems or departments.
This doesn't need to result in another layer of administration. It may mean introducing information governance requirements into existing AI approval processes, supplier assessments and risk reviews.
The important thing is that information governance becomes part of AI decision-making before access is granted, rather than a problem to address afterwards.
Six questions Legal should ask before corporate information is made available to AI.
You don't need to solve every information governance problem across the business before AI can be used. But you do need confidence that the information within the proposed deployment has been assessed and that the relevant legal risks are understood.
These six questions provide a practical starting point.
The answers may identify gaps requiring further investigation or action. Some can be addressed through existing processes, while others may require changes to information governance policies, responsibilities or technology.
What matters is that these decisions are made deliberately, with the right stakeholders involved.
1. What information will AI be able to access, and who is responsible for it?
Identify the repositories, categories of information and business owners involved. Establish whether information is subject to confidentiality, privilege, contractual restrictions or other legal obligations.
2. Are existing access restrictions appropriate and enforceable?
Work with IT and Security to confirm that users can only access information they are authorised to see. Review any restrictions that should apply to particular information categories or uses.
3. Are retention and defensible disposition requirements being followed?
Establish whether information has been retained appropriately and whether material that is no longer required has been reviewed for disposal. Consider applicable legal holds and requirements affecting downstream copies or AI indexes.
4. Can preservation requirements be maintained?
Review whether legal hold and investigation procedures account for information held or generated within AI-enabled workflows. Establish how preservation instructions will be communicated, implemented and evidenced.
5. Have the legal and contractual implications of AI processing been assessed?
Consider data protection, confidentiality, supplier obligations, third-party processing and any relevant restrictions on how information can be used, stored or disclosed.
6. Who is accountable for maintaining these controls?
Agree responsibilities for approving access, implementing technical safeguards, monitoring compliance, handling exceptions and reviewing requirements as the AI deployment expands.
Information governance should enable AI adoption, not hold it back.
It's easy to position information governance as another set of requirements that must be satisfied before an AI project can proceed.
But that misses its wider value.
When you understand what information your business holds, who owns it, what restrictions apply and how it should be retained or disposed of, you are better placed to make informed decisions about how AI can use it.
You can identify information that is appropriate for a particular AI use case, establish the controls needed to protect sensitive material and reduce unnecessary exposure. Clear governance also makes it easier to expand AI adoption because responsibilities and processes have already been defined.
For in-house legal teams, this creates an opportunity to contribute to AI adoption in a constructive way. Rather than being seen solely as the function identifying potential problems, Legal can help establish the conditions under which the business can use information responsibly and with greater confidence.
This doesn't mean information governance alone will make AI safe or compliant. AI-specific risk assessments, technical safeguards, testing and oversight remain essential.
But without appropriate control over the underlying information, those wider safeguards may not be sufficient.
From AI ambition to legal accountability.
AI is changing how corporate information is accessed, processed and used. For in-house legal teams, that makes familiar obligations around confidentiality, privilege, retention, preservation and accountability more important than ever.
The challenge isn't to prevent the business from adopting AI or to make Legal responsible for every aspect of its implementation.
It's to ensure that the legal obligations attached to corporate information remain understood and enforceable as technology changes.
That requires cooperation across departments, clear ownership and practical information governance controls that can be applied consistently across existing systems.
By becoming involved early, asking the right questions and helping establish shared accountability, Legal can support responsible AI adoption while reducing unnecessary information-related risk.
AI may change how your business uses information. It doesn't change your responsibility to ensure that information is appropriately governed.
Put information governance into action.
iCompli helps you understand and manage information across your existing systems, apply consistent retention and disposition policies, coordinate governance workflows and maintain evidence of decisions.
Reduce information risk, control unnecessary costs and turn governance requirements in to practical action.
Sources
1. Association of Corporate Counsel — AI in Legal Practice: What In-House Counsel Must Know in 2026
May 2026
Supports the discussion of privilege, discovery, professional responsibility and the legal considerations surrounding generative and agentic AI.
onlineed.acc.com
2. Solicitors Regulation Authority — Misuse of AI: Warning Notice
August 2026
Particularly relevant to in-house solicitors. Addresses confidentiality, privilege, the risks of using AI tools and the need for appropriate safeguards.
Solicitors Regulation Authority
3. Information Commissioner's Office — Guidance on AI and Data Protection
Supports the discussion of personal data processing, minimisation, retention and security when AI systems are used. The ICO notes that parts of this guidance are under review following the Data (Use and Access) Act.
ICO+1
4. National Cyber Security Centre — Guidelines for Secure AI System Development
Supports the arguments around access controls, AI information processing, secure deployment and ongoing monitoring.
National Cyber Security Centre