For me 2024 is going to be the year, I try to scale the information security compliance consulting I have been on and off doing for last 20+ years into a combination of AI models and tooling. This is why this topic feels close to heart. The article will also be published in my company's www.peakdefence.com page once we launch the new one!
One has to start with a short reflection into history of information security compliance to understand where it might be going in the future! In short there have been attempts to find ways to create a standard “cookbook” which would allow to achieve information security while following a predefined sequence of actions / list of activities.
In short the reasons of compliance temptation have been summarised in table below based on two main groups of stakeholders.
Lets define two potential sides within the compliance implementation. Let’s call one side Foxes and the other one Rabbits, where the Foxes define the requirements while Rabbits try to navigate them while staying alive and healthy.
When applying these roles and looking at history of information security compliance, one can see the role of Foxes is often taken by government agencies or larger corporations while Rabbit role is on the organizations willing to do business with Foxes like startups or scale-ups.
As a result the current landscape in information security compliance looks like this:
there are a few widely recognised standards which organisations can certify against and the choice is usually defined by industry and geography combination (e.g. a SaaS provider in EU with most customers in EU would likely want to certify with ISO 27001 or ISAE 3000 (if for example in Denmark) while one in US or with US larger customers would be more concerned with SOC 2 or similar)
All of them follow similar principles of setting security objectives, identifying and managing risks, implementing a more or less predefined (depending on standard) structure of controls and having reviews and audits on regular basis
Very often one can observe major struggles in implementing compliance programs efficient and effective even in large Foxes (the ones with many resources, etc) with a lack of understanding among internal stakeholders “why we do this”
Small Rabbits often go for compliance just to land the bigger “customer” without understanding the real reasons and assistance compliance program can provide. This more often than not leads to “unbalanced” compliance programs which do just one thing (e.g. secure configurations) and claiming “all work is done”
As an outcome of the factors mentioned above following can be observed:
Certified “to the teeth” companies still end up having security incidents
There is a lot of misunderstanding out there what certification means and what scope means and how it can/should be applied in supply chain
Growth in tooling which provides “checklist driven certification”
From my perspective the summary is very simple. Effective information security is “balanced” = takes into account (A) People, (B) Processes and (C) Technology on somewhat “even” level! If even one of those fail, it might quickly lead to much bigger failure!
As many companies implementing information security programs come from tech, our the initiatives to do compliance are often perceived as “tech”, many company management systems end up heavily Technology biased.
Also one more critical role here is the auditors and their background and competence! While working with many auditors as trainer or lead auditor I have observed very personal background driven impact on auditors decisions on what to audit, how to audit and what the findings are!
Current generative AI based on the challenges the industry is facing now will probably have a huge impact on the industry as a whole. It is both expected to reduce the
Key process: checking if Rabbits “fall in line” with information security.
Key process 1: demonstrating compliance to Hedgehogs (auditors) an Foxes (customers, partners)
Process element:
Answering RFPs, DD questions, etc
How it is done now:
Manual work collecting the necessary information.
How AI might change it?
Positive:
- Major reduction in time spent
- Better and more consistent answers
Negative:
- Biased towards “positive” answers
Implementing really effective and compliant management system
Lets add one more role here. Let’s call them hedgehogs. These would be the companies and individuals dealing in providing assurance everything “is good enough”. One of the main reasons for them to exist is the lack of direct trust between Foxes and Rabbits, but also an option to review how good the expectations of certain compliance requirements are met by spending less (no Fox going to all Rabbit holes to check and not all Rabbits being happy about regular visits from different Foxes).
Key process: auditing if compliance is done properly and obtaining reasonable assurance
Existing processes for compliance checking are quite time, knowledge and resource intensive and quite often involve repetitive tasks.
There is a huge potential for AI at least in following areas of compliance improvement if used properly
Policy and Procedure creation support
Effective and balanced implementation of compliance management can become easier if proper tooling is used (the one which combines organizational knowledge with security domain knowledge).
Specific policies and procedures which are not templates but driven by model’s knowledge of the organization
Information collection support
Security event, vulnerability, threat, incident and risk identification areas
Supporting processes to identify suspicious events from conversations or other areas
Knowledge support
Making information available easily (much better google to find information internally)
Awareness and Training support - Ensuring the employees are aware in specific areas and keeping up their profile
Avoiding duplicates an contradictions in the information (AI can find and indicate those in the documentation)
Audit support
Stage 1 automated audit
Stage 2 audit planning automation
Stage 2 audit support by indicating what evidence to collect and how
There is also large set of risks to abuse AI to demonstrate “virtual” (not actual) compliance, where organization can start answering questions “correctly” event without having a proper system in place