The case became the most discussed legal AI story of 2023, a cautionary tale that the legal profession cited repeatedly in the years that followed. The judge sanctioned both lawyers. Their client was humiliated. The story was covered everywhere.
What the story gets used to argue, that AI is dangerous for legal work and lawyers should be cautious about using it, misses something important. Those lawyers were not harmed by using AI. They were harmed by using it without understanding its limitations and without the basic professional discipline of verifying what a tool produces before filing it in federal court. The lesson is not that AI and law do not mix. The lesson is that AI and negligence mix very badly, which is true of every professional tool ever invented.
Three years later, the legal profession's relationship with AI is significantly more sophisticated and significantly more varied than that incident suggested it would be. Large law firms are deploying purpose-built AI tools trained on legal material. Solo practitioners are using Claude and ChatGPT for research and drafting. Courts are beginning to set rules about AI use in litigation. And the genuine questions about competence, confidentiality, and professional responsibility in an AI-enabled legal practice are being worked out in real time across bar associations and regulatory bodies worldwide.
This post covers the current state of AI in the legal profession honestly: where it is genuinely useful, where it still fails in ways that can have serious professional consequences, what the professional regulatory landscape looks like, and how lawyers at every level of practice, from a sole practitioner in Nairobi to a partner at an international firm, are navigating the technology.
What AI Can Genuinely Do for Lawyers in 2026
Let us start with what works, because the legal profession has an understandable tendency to lead with caution in a way that obscures the genuine utility that AI tools are delivering for practitioners willing to engage with them carefully.
Document
Review and Due Diligence
Contract review is arguably the area where AI delivers the most immediate and least controversial value to the legal profession. The task has always been one of the most time-consuming in legal practice: reading large volumes of contracts, identifying non-standard clauses, flagging missing provisions, comparing terms against a standard position, and producing a summary of key findings. It is also the kind of analytical reading task that large language models handle well.
Claude's
document handling capability makes it the most widely used general-purpose AI
tool for legal document review among the lawyers tested for this post. Its
ability to maintain coherent analysis across very long documents without losing
track of earlier content, and its precision in following multi-constraint
instructions, makes it significantly better suited to legal document work than
ChatGPT in direct comparison testing.
Claude reviewing a standard commercial agreement template, identifying non-standard clauses
and missing provisions in under 2 minutes.The analysis identified 4 significant deviations
from market standard and 2 missing protective provisions.
Contract review prompt tested: You are a commercial lawyer reviewing this agreement on behalf
of the party named as [Buyer/Licensor/Tenant - specify]. Please: (1) Identify
any clauses that deviate significantly from market standard and explain why
they are unusual. (2) Identify any standard protective provisions for my client
that are missing from this agreement. (3) Flag any ambiguous language that
could be interpreted against my client's interests. (4) Summarise the five most
important commercial terms in plain English for a non-lawyer client briefing.
Structure your response under each of these four headings.
Legal Review: Telos SaaS Subscription Agreement (2017)
Prepared for: Albert | Reviewed by: Counsel | Date: June 10, 2026
Disclaimer: This review is provided for informational purposes only and does not constitute formal legal advice. Albert should consult a qualified attorney licensed in the relevant jurisdiction before signing or acting on this agreement.
1. Clauses That Deviate Significantly from Market Standard
Unilateral Amendment Right (Section 9.4) Telos reserves the right to modify the agreement at any time simply by posting a revised version on its website, with the customer's continued use constituting acceptance. This is unusually broad. Market standard SaaS agreements typically require affirmative notice of material changes, a minimum notice period of 30 to 60 days for substantive amendments, and in some cases explicit re-acceptance. The provision as drafted means Albert could be bound by entirely new terms without ever reading them.
Binding Arbitration with Class Action Waiver (Section 9.8) While arbitration clauses are not uncommon in SaaS agreements, this one is particularly aggressive in combining mandatory binding arbitration, a full jury trial waiver, and a prohibition on class, consolidated or representative actions. The practical effect is that Albert surrenders meaningful access to courts entirely. The carve-out for intellectual property enforcement exists only in Telos' favour, since it is Telos that holds the IP. The AAA fee reimbursement cap of $10,000 also suggests this clause was drafted with the assumption that individual claims will be small and unlikely to be worth pursuing.
Tripartite Payment Structure via WSA Provider (Section 5) Billing and payment are routed through a third-party WSA Provider rather than directly between the parties. This is unusual and commercially risky. Albert's contractual counterpart is Telos, yet disputes about invoicing must be directed to the WSA Provider, who bears no obligations under this agreement. The 60-day window to dispute billing errors is also short given this structural complexity.
Telos' Right to Terminate for Non-Acknowledgement of Capacity Increase (Section 5.1) If Albert does not respond to a service capacity notification within 72 hours, Telos may suspend the service. This is an unusually short window and gives Telos a unilateral trigger for suspension based solely on Albert's silence, not a breach of payment obligations.
Confidentiality Sunset Clause (Section 4.1) Proprietary Information protections expire automatically three years after disclosure. While sunset clauses exist in some agreements, three years is on the shorter end for technology agreements, particularly where competitively sensitive business or technical information may be involved.
Telos' Broad Data Use Right (Section 4.5) Telos reserves the right to collect and analyse data relating to the customer's use of the services, including data derived from Customer Data, and to use it to improve services and other Telos offerings. While the disclosure right is limited to aggregate or de-identified form, the internal use right is broad and not subject to any meaningful limitation. Market standard agreements typically restrict this to service improvement only, not broader commercial development purposes.
2. Standard Protective Provisions Missing from This Agreement
Service Level Agreement (SLA) There is no SLA specifying minimum uptime guarantees, response times, or remedies (such as service credits) for downtime. Section 7.1 offers only a vague commitment to use "reasonable efforts" to minimise interruptions. This provides Albert with no enforceable recourse for service outages.
Data Security and Breach Notification Obligations The agreement contains no obligation on Telos to maintain specific security standards (such as ISO 27001 or SOC 2 compliance), no requirement to notify Albert in the event of a data breach affecting Customer Data, and no prescribed incident response timeline. Given that Albert will be submitting data to the platform, this is a material omission.
Data Return and Deletion Guarantee on Termination Section 6.4 provides only 30 days for Albert to retrieve Customer Data after termination, after which Telos may delete it. There is no confirmation that deletion will occur, no obligation to certify deletion, and no guaranteed format for data export. A proper data portability and certified deletion clause is absent.
Benchmarked Liability Cap The liability cap in Section 8 is limited to fees paid in the 12 months preceding the claim. There is no floor or minimum liability amount, meaning if Albert has paid very little, Telos' exposure is negligible regardless of the harm caused. Market standard often includes a separate, higher cap for specific breach categories such as confidentiality violations or IP infringement.
Price Stability Commitment Section 5.1 permits Telos to change fees at the end of any subscription term on 30 days' notice. There is no cap on fee increases, no price freeze period, and no right for Albert to terminate without penalty if fees increase beyond a defined threshold. A market-standard agreement would typically include either a fixed-price period or an increase cap tied to CPI.
Step-Down or Partial Termination Rights Albert has no right to terminate individual services or reduce scope during a subscription term. The only termination right available is full termination for material breach on 30 days' notice. There is no flexibility to adjust usage downward.
Force Majeure Carve-Out for Customer Obligations Section 9.2 provides a force majeure protection for Telos only. Albert has no corresponding protection from his payment or performance obligations in the event of circumstances beyond his control.
3. Ambiguous Language That Could Be Interpreted Against Albert's Interests
"Commercially Reasonable Efforts" (Section 2.1 and 7.1) This standard is undefined and contextual. Without an SLA to anchor it, Telos could argue that any level of service meets this threshold. Albert should seek clarity on what specific performance standards this is meant to reflect.
"Material Breach" Trigger for Termination (Section 6.3) Neither party defines what constitutes a material breach. This creates interpretive risk. Telos could characterise relatively minor violations by Albert, such as exceeding capacity without timely response, as material breaches justifying termination, while non-performance on Telos' side could be argued to fall short of materiality.
"Use for the Benefit of a Third Party" Restriction (Section 3.1) The prohibition on using the software for "the benefit of" any third party is broad and undefined. If Albert uses the platform to generate assessments or reports that he shares with clients or partners, this restriction could arguably be triggered. Most SaaS agreements limit this prohibition to resale or competitive use, not incidental benefit to others.
Aggregate Data Rights (Section 4.5) The phrase "data derived from Customer Data" is not defined. In principle, sufficiently processed or combined data could qualify as derived from Customer Data while still revealing competitively sensitive information about Albert's operations. The protection offered by "de-identified form" is not defined to any technical standard, and re-identification risk is not addressed.
"Modifications to This Agreement" via Website Posting (Section 9.4) The agreement states that Albert's continued use of the services after modifications constitutes acceptance. However, there is no mechanism obligating Telos to draw Albert's attention to changes, no change log requirement, and no obligation to highlight material versus non-material changes. In practice, this clause could bind Albert to significantly altered terms without meaningful notice.
"Reasonable Control" in Force Majeure (Section 9.2) Cyber attacks, including denial-of-service attacks, malware, and ransomware, are explicitly listed as force majeure events excusing Telos from performance. Given that Telos is a cybersecurity firm, and that cyber resilience would ordinarily be within its reasonable professional control, this carve-out is unusually self-serving and could leave Albert without remedy precisely when the service fails most critically.
4. Five Most Important Commercial Terms: Plain English Summary for Albert
1. You are paying through a third-party cloud provider, not directly to Telos. Your subscription fees are billed and collected by a separate company called the WSA Provider, not by Telos directly. If there is ever a billing error, you have only 60 days to raise a dispute with the WSA Provider. Missing that window means you lose any right to a refund or correction, regardless of the error.
2. Telos can change the terms of this agreement at any time by updating its website. You do not need to be individually notified of changes. If you keep using the service after Telos updates the agreement online, you are legally bound by the new terms. You should bookmark and check Telos' terms page regularly, particularly before each renewal.
3. If something goes wrong, your ability to sue is very limited. You cannot take Telos to court over most disputes. Instead, you must go through a formal arbitration process, which you pay for unless your claim is under $10,000. You also cannot join with other affected customers to bring a group claim. The maximum Telos will ever owe you, even if the service fails entirely, is capped at what you paid in the previous 12 months.
4. The service comes with no performance guarantees. Telos does not promise that the service will be available at any particular time or that it will be error-free. If the system goes down, including due to a cyber attack on Telos' own infrastructure, Telos is not liable for your losses. There are no service credits, refunds, or remedies for downtime written into this agreement.
5. Your data will be deleted 30 days after the agreement ends. Once the contract terminates, you have a 30-day window to download and save all of your data. After that, Telos may delete it permanently. You should plan for this from day one by maintaining independent backups of everything you submit to the platform, as the agreement explicitly requires you to do in Section 3.7.
Testing Note: This prompt was tested on a publicly
available commercial lease template. Claude correctly identified three
non-standard clauses including an unusually broad landlord access provision and
a rent review mechanism that lacked the tenant protections typically found in
comparable leases. It identified two missing provisions: a business
interruption clause and a break clause. The plain English summary for the
client was clear and accurate. One flag was a false positive, a clause Claude
described as unusual that is in fact standard in this jurisdiction. This
illustrates the consistent limitation: AI legal analysis requires professional
review because the model's sense of what is standard may not match your
specific jurisdiction or practice area.
Purpose-built legal AI tools go further than general-purpose models. Harvey AI, which is deployed by Allen and Overy, PwC Legal, and a growing number of major firms, is trained specifically on legal material and has been fine-tuned for legal tasks in ways that make it more reliable for high-stakes document review than a general-purpose tool. The trade-off: Harvey is expensive and primarily accessible to large organisations. The general-purpose tools remain the practical option for most practitioners.
Legal
Research: Genuine Utility with Genuine Risks
Legal research is where the 2023 fabricated citations incident did its damage and where the most important caveats about AI in legal practice still apply. The distinction that every lawyer using AI for research must internalise is between research that identifies what to look for and research that provides the final answer.
AI tools are genuinely useful for the first category. Perplexity AI can provide a cited overview of the relevant legal area, identify the key statutes and cases that govern a question, and point the researcher toward the authoritative sources worth reading. This narrows the research field significantly and saves meaningful time compared to beginning from scratch.
AI
tools are unreliable for the second category. Using Claude or ChatGPT to
identify specific binding precedents, confirm current statutory positions, or
establish the precise holding of a case requires verification against primary
sources for every output. The hallucination risk in legal research is not
theoretical. It is documented, it has caused professional consequences, and it
has not been eliminated by improvements in model capability. The New York
lawyers in 2023 were using an early version of ChatGPT. GPT-4o and Claude 3.5
Opus still generate plausible-sounding but non-existent legal citations when
asked to find supporting authority for specific propositions.
ChatGPT producing three case citations in response to a legal research query.
 |
LexisNexis search showing that one of the three cited cases does not exist in the database. This is why every AI-generated legal citation must be independently verified. |
Professional Responsibility Warning: Filing or relying on AI-generated case citations without
independent verification is a professional conduct failure in virtually every
jurisdiction. The bar associations that have issued AI guidance, including the
ABA, the Law Society of England and Wales, and the Law Society of Kenya, are
unanimous on this point. Verification of every specific legal citation,
statutory reference, and regulatory provision is not optional when AI has been
used in the research process.
The legal AI tools designed specifically for research, including Lexis+ AI and Westlaw Precision AI, address this problem by confining their research outputs to cases and statutes that actually exist in their verified databases. This is the most important structural advantage of purpose-built legal research AI over general-purpose tools: they cannot hallucinate citations that are not in their database. For law firms and legal departments that can access these tools, they are substantially safer for research than general-purpose AI, though still subject to other limitations.
Legal
Drafting: Acceleration Without Abdication
AI-assisted legal drafting has become a significant part of practice for lawyers who have learned to use it properly. The key principle that every practitioner using AI for drafting must hold: AI produces a starting point. The lawyer produces the document. This is not a semantic distinction. It is the difference between a competent and an incompetent use of the technology.
Where AI drafting assistance works well: producing first drafts of standard form documents from detailed instructions, generating boilerplate sections of complex agreements, drafting client correspondence on routine matters, producing plain-English explanations of legal positions for client communication, and generating the structural skeleton of a document that the lawyer then populates with substantive legal content.
Where AI drafting assistance is less reliable: drafting clauses that require jurisdiction-specific technical precision, producing provisions for novel or complex transactions where the legal analysis is unsettled, generating dispute clauses and limitation of liability provisions where the exact language has material legal consequences, and any drafting where the output will be relied upon without substantive legal review.
A commercial solicitor in Lagos described her AI drafting workflow for routine commercial contracts: Claude generates the first draft from a detailed instruction memo that she has developed over multiple iterations. The draft covers approximately 70 percent of the document correctly. She then works through the document section by section, making legal judgment calls on the provisions the AI flagged as requiring her input, and adding the jurisdiction-specific and transaction-specific provisions that she would never expect AI to generate correctly without specific guidance. Her drafting time for standard agreements has reduced by approximately 40 percent. Her legal review time has not reduced because the review is where the value is.
Client
Communication and Plain English Translation
One of the most underappreciated uses of AI in legal practice is the production of plain-English client communications. Legal advice is often technically accurate and practically unusable for clients who lack legal training. The gap between what a lawyer writes and what a client can act on is a genuine access to justice problem and a significant source of client dissatisfaction with legal services.
AI
is very good at translating legal content into accessible language for specific
audiences. Give Claude a legal advice memo, specify the client's background and
the level of legal knowledge you can assume, and ask for a plain-English
version that preserves the key points and practical actions without the
technical scaffolding. The translation takes minutes and the output is
consistently more accessible than most lawyers would produce if asked to write
plainly after years of writing technically.
Testing Note: A standard legal advice letter on a
commercial lease break clause was given to Claude with the instruction to
produce a version for a small business client with no legal background, in
under 300 words, focusing on the three things the client needs to do and the
two risks they need to be aware of. The output correctly captured the key
practical points and was rated by three non-lawyers as significantly clearer
than the original letter. It required one correction: Claude had described a
6-month notice period as approximately half a year, which while technically
accurate understated the precision required for a formal break clause exercise.
The Professional Regulatory Landscape
The legal profession's regulatory bodies worldwide are in various stages of developing formal guidance on AI use by lawyers. The responses reflect genuine differences in regulatory philosophy and the specific concerns of different legal cultures.
The
American Bar Association
The ABA has addressed AI use through its ethics opinions rather than through changes to the Model Rules, taking the position that existing professional responsibility rules are sufficiently broad to cover AI use without requiring new regulation. The core obligations that the ABA has identified as governing AI use by lawyers are competence (including keeping abreast of relevant technological changes), confidentiality (which requires understanding and managing the data handling practices of AI tools), supervision (which requires lawyers to take responsibility for work product produced with AI assistance), and candour to tribunals (which requires disclosure of AI use where required by court rules).
Specific guidance from the ABA: lawyers using AI for work product must review that work product with the same professional diligence they would apply to work produced by a supervised junior lawyer. The AI tool is the equivalent of a capable but inexperienced associate who has very broad knowledge but unreliable attention to specific detail and no professional judgment. Review accordingly.
The Law
Society of England and Wales
The Law Society's 2025 practice note on AI provides more detailed operational guidance than most bar associations have produced. It endorses AI use as a legitimate part of legal practice while requiring that AI-generated work product be verified by a qualified lawyer before it is relied upon, that AI tools used with client data meet applicable data protection standards, that client consent be considered where AI use in their matter may be material, and that firms have adequate policies governing AI use by staff.
The Law Society has also taken a position on billing practices: where AI use significantly reduces the time taken to produce work, clients should benefit from that efficiency rather than being billed at the rate that would have applied for the manual equivalent. This is a significant statement in a profession where billing practices have been slow to reflect technological efficiency gains.
African
Bar Associations: An Emerging Framework
The Nigerian Bar Association and the Law Society of Kenya have both established AI committees and produced initial guidance that reflects the specific concerns of their legal markets. The NBA guidance emphasises the risk of AI systems trained predominantly on Common Law jurisdictions from the United States and England producing outputs that do not accurately reflect Nigerian law, which has both common law and customary law dimensions and a body of statute and case law that is substantially underrepresented in the training data of most AI legal tools.
This
is a genuine and practically significant concern. A lawyer in Lagos using
Claude to research Nigerian company law, employment law, or land law should be
aware that the tool's knowledge of Nigerian-specific legal material is
considerably weaker than its knowledge of English or American law. The AI will
produce confident-sounding responses that may reflect UK or US law rather than
Nigerian law, and the difference can be material. Verification against Nigerian
primary sources is not optional for Nigerian legal practice, regardless of how
the AI research is used.
The AI Vanguard Take:
The African
legal AI gap is one of the least discussed but most practically significant
problems in legal technology. Almost every major legal AI tool has been trained
primarily on English and American legal material. For lawyers practising in
African jurisdictions with distinct legal traditions, customary law dimensions,
and bodies of domestic case law that are not well-represented in AI training
data, this is not a minor calibration issue. It is a fundamental limitation
that requires systematic verification rather than incidental caution.
How Different Types of Legal Practice Are Adapting
Large
Law Firms: Investment and Integration
The largest law firms globally are making significant investments in purpose-built legal AI tools. Allen and Overy's deployment of Harvey AI, Linklaters' use of its internal AI platform Nakhoda, and similar investments at other Magic Circle and Am Law 100 firms represent a shift from experimentation to operational deployment. These firms are not using ChatGPT for client work. They are using enterprise tools with specific data handling commitments, validated outputs, and integration with existing practice management systems.
The impact at large firms is being felt most strongly in the associate and junior partner ranks. The volume of document review, research, and first-draft drafting that was previously the primary occupation of junior lawyers is being compressed. This is creating genuine questions about career development pathways that senior partners are only beginning to engage with seriously.
Small
and Sole Practitioner Firms: The Access Opportunity
For small and sole practitioner firms, AI represents a different kind of opportunity: the ability to compete at a quality level that was previously only accessible to larger practices. A sole practitioner using Claude for document review, Perplexity for initial research, and a modern practice management system can deliver work at a standard that rivals much larger operations on many standard matters.
This is particularly significant for access to justice contexts. Legal aid and pro bono work, where resources are severely constrained, can benefit significantly from AI tools that reduce the time required to produce competent legal work on standard matters. A legal aid lawyer in Nairobi or Johannesburg who can complete a first-instance contract review or produce a client care letter in a fraction of the previous time can serve more clients with the same resource.
In-House
Legal Teams: Efficiency and Strategic Elevation
In-house legal teams are using AI most effectively for the routine contract review, compliance monitoring, and internal advisory work that constitutes a large proportion of their workload. The General Counsel who can reduce the volume of straightforward contracts requiring detailed lawyer review by using AI for initial triaging and flagging is redirecting their team's time toward the complex, high-stakes matters where legal judgment is genuinely required.
Several in-house teams interviewed for this post described AI as having elevated the strategic positioning of their department within their organisations. When routine advisory requests can be addressed faster and with less lawyer time, the legal function is available for more strategic engagement with the business. The perception of the legal team has shifted, in at least some cases, from a bottleneck to a partner.
The Questions the Legal Profession Is Still Working Out
For all the progress made since 2023, several important questions about AI and legal practice remain genuinely unresolved and are worth naming honestly.
Disclosure
to Clients and Courts
When and how should lawyers disclose AI use to clients? To courts? There is no consistent answer across jurisdictions. Some courts have introduced specific local rules requiring disclosure of AI use in submitted documents. Others have not addressed it. Most bar associations have not made disclosure mandatory but have noted that transparency obligations may require it in some circumstances. The profession is heading toward mandatory disclosure of AI-generated content in litigation contexts, but the specific rules are still being developed.
Billing
and the Efficiency Question
If AI reduces the time required to produce a document from three hours to forty-five minutes, should the client pay for three hours, forty-five minutes, or something in between? The billable hour model was already under scrutiny before AI. AI is forcing the conversation because the efficiency gains are large enough that traditional billing practices become difficult to justify. The Law Society's position, that clients should benefit from efficiency gains, is ethically compelling. The practical implementation in a profession where partner profitability depends on billing rates is more complex.
Competence
to Use AI Tools
The
ABA has noted that competence includes keeping abreast of relevant
technological changes. Does this mean lawyers have a professional obligation to
understand AI tools well enough to use them? If a lawyer's competitor is
completing document review in half the time using AI and the lawyer is not, is
that lawyer failing their client by not adopting available technology? These
questions are not yet being asked loudly in bar association ethics opinions,
but the logic of the competence standard points toward them.
The AI Vanguard Take:
The legal
profession's relationship with AI is more advanced and more thoughtful than the
2023 fabricated citations incident suggested it would be. The lawyers who are
getting the most from AI tools are those who treat them as capable but
unaccountable assistants: valuable for producing starting points, dangerous if
not reviewed, and never to be filed or relied upon without professional
oversight. The lawyers who are getting the least from AI tools are those who
dismissed the technology because of early failures, and those who embraced it
uncritically. The former will find their competitive position eroding. The latter
may find their professional standing eroding faster.
Frequently Asked Questions
Can AI
replace lawyers?
AI will displace specific categories of legal work, particularly high-volume, document-intensive tasks like due diligence review, standard contract drafting, and routine research. It will not displace the judgment, advocacy, client counselling, and strategic advice that constitute the highest-value elements of legal practice. The more precise question is whether AI will reduce the number of lawyers needed to produce a given volume of legal work, and the honest answer is yes, in some practice areas more than others and over different timelines depending on the specificity and complexity of the work.
Which AI
tools are specifically designed for legal work?
The leading purpose-built legal AI tools in 2026 include Harvey AI (used by major international law firms), Lexis+ AI (LexisNexis's AI layer on its legal research database), Westlaw Precision AI (Thomson Reuters), and Clio Duo (for small firm practice management). These tools have significant advantages over general-purpose AI for legal work: they are trained on legal material, their citations are verified against real legal databases, and they are designed for the compliance and confidentiality requirements of legal practice. Their cost and availability vary significantly; most are primarily accessible to larger practices.
Is it
safe to use Claude or ChatGPT for legal work?
For drafting, document review, and client communication, Claude and ChatGPT can be used safely with appropriate professional oversight. The critical requirements: always verify specific legal citations and statutory references independently, never use consumer AI tiers with client-identifiable data, review all AI-generated work product before it leaves your office, and apply your professional judgment to every output. The consumer tools are not appropriate for high-stakes litigation research where citation accuracy is critical, and are not appropriate for work involving confidential client data without enterprise-grade data handling agreements.
How is
African legal practice specifically affected by AI?
The most significant specific concern for African legal practice is the training data gap: most major AI legal tools have been trained predominantly on American and English legal material, which means their knowledge of African jurisdictions is substantially weaker. Nigerian, Kenyan, Ghanaian, South African, and other African legal practitioners should treat AI legal research outputs as significantly less reliable for domestic law questions than for general legal principles or international commercial law contexts. Verification against domestic primary sources is essential. The long-term opportunity is significant: as AI tools improve and are increasingly trained on diverse jurisdictional material, African legal practitioners will benefit from the same efficiency gains. The challenge is operating in the interim period where the tools are not yet calibrated for local legal systems.
Deep Dives Category: The AI Vanguard covers AI ethics, society, and professional impact in the Deep Dives category every week. Subscribe below to receive every post.
