What managing partners should ask AI vendors before signing any contract
In this Help Net Security interview, Kumar Ravi is the Chief Security & Resilience Officer at TMF Group, argues that over-privileged access and weak workflow controls pose more danger than ransomware attacks, precisely because they accumulate quietly and go unnoticed.
He addresses the tension between legal privilege and timely threat-sharing, the challenge of managing fourth-party vendor risk, and the questions firms should ask before adopting AI-native tools. His broader argument: security needs to sit at board level, measured consistently and backed by independent assurance.
The breach that keeps you up at night isn’t always the dramatic ransomware attack. In your experience, what’s the most underestimated threat vector specific to professional services, and why does it keep getting overlooked?
Two of the most underestimated threat vectors are gradual and imperceptible. These are over-privileged accesses, for which users have more permissions than necessary for their roles, and weak controls in everyday workflows.
These lead, slowly and unnoticeably, to a loss of confidentiality and data protection. There is no need for a one-time-off “big bang” data breach, but these small breaches in access and privileges that happen every day are even more dangerous. These can be found in access to applications, shared service accounts in document systems, etc.
This tends to go overlooked because it rarely explodes into an immediate crisis. Instead, it is often seen as an apparently innocent internal policy violation.
However, these small oversight errors accumulate across multiple processes, teams, systems and applications. When one aggregates them, they can lead to an attack opportunity for lateral movements for insiders and outsider threats.
This is particularly dangerous if governance is not in place. If no one is responsible for overseeing this, no one really owns it. Thus, putting data governance in the centre, and focusing on identity and access management by design, is key to preventing risks and being prepared to mitigate them if they appear.
Privilege and confidentiality protections create this strange dynamic where firms are reluctant to share threat intelligence with each other or even with regulators. How much does that legal culture actively work against collective defence?
Legal privilege and confidentiality are essential, but the challenge is that, under pressure, companies increasingly tend to treat all data points as privileged. This can be counterproductive, as it can slow down and jeopardise timely information-sharing.
This is an important challenge, as regulators and peers require speedy, specific and actionable insights.
Companies need to find a balance between privilege and confidentiality protections, and a strategy for sharing threat intelligence. Done well, this can help improve the overall ecosystem resilience without compromising our legal culture.
Small and mid-sized firms outsource enormous amounts of sensitive work, from document review to financial modeling to cloud infrastructure. How should a firm be thinking about third-party risk when their vendors’ vendors also have access to client data?
Vendors should no longer be seen as an external third party. They should become part of a company’s integrated supply chain, with the duties, responsibilities and trust that this requires.
At a practical level, the company’s responsibility is to understand the end-to-end supply chain and the role that every vendor plays in it. This means, for example, keeping a live inventory of all partners that handle sensitive data, classifying them and even requiring vendors to disclose their sub-processors.
When signing a contract, companies need to set security obligations, incident-notification timelines and the right to know who accesses their data and in which terms. Additionally, due diligence needs to go beyond a one-off questionnaire and needs to be assessed periodically. You need to make sure that you set minimum standards and that your vendors enforce them.
You can outsource work, but you can not outsource the accountability to protect your and your client’s data. So, you need to do everything in your power to enforce your standards.
Legal tech and fintech have exploded with AI-native tools promising efficiency gains. What questions should a managing partner be asking before signing a contract with any of these vendors, and which answers should be dealbreakers?
Onboarding AI-native tools should be seen as giving new hires access to firm’s most sensitive information. It needs to be done by giving specific guidelines, and without making any assumptions.
Specific questions need to be made: What does the tool take in, where is it based and which jurisdictions’ laws apply to it? Is the information used to train models that benefit other customers or is it kept strictly to that company’s environment? What are the retention rules, what are the processes to request deletions and what audit logs are in place?
On top of that, independent assessments to track these are required. These can be audit reports, certifications, pen-test summaries, incident records…
When contracting, companies also need to be very specific about accountability. They need to be clear on what they commit regarding security controls, breach notification timelines and liability if a failure leads to financial, regulatory or reputational damage.
If you could change a structural thing about how professional services firms are built, governed, or regulated to make them more secure, what would it be and who would resist it most?
If I could make a structural change, it would be to treat security as a core business control, owned at the board level, measured consistently and evidenced through independent assurance. The aim is to work towards a predictable resilience. This means having fewer points of failure, faster containment when something goes wrong and more transparency.
Download: CIS Benchmarks March 2026 Update