Which framework, and what to build first
The starter kit most startups put in place first, then SOC 2 and the frameworks that sit around it, and a way to choose. (About 14 minutes)
Earlier: Do you need it yet?. This part stands on its own if you’d rather start here.
Your Compliance Starter Kit
Here is a list of measures most startups put in place first, regardless of which framework they’re going after. Use it as a simple starting point. This is not an exhaustive list—your auditor may expect more, depending on your industry, product, team size, and infrastructure.
Your ticks stay in this browser only. They’re there to help you see where you stand today.
Frameworks you’ll most likely encounter
If you’re a tech company selling in the US, SOC 2 is almost certainly the first framework you’ll encounter. There’s a reason we’ve mentioned it so many times—SOC 2 is the default trust signal customers expect to see when they ask whether you’re secure. So let’s start there.
The essentials of SOC 2
What it is: SOC 2 is a framework created by the American Institute of Certified Public Accountants (AICPA), the US accounting body.
The framework is interpretive rather than prescriptive, which means it lists what you need to achieve, not what controls you implement. So you can design controls that you think best fit your business. A Certified Public Accountant (CPA) or a CPA firm then reviews these measures and compiles a report of their opinion and findings.
Who it applies to: SOC 2 is an attestation and applies to service organizations—any company that stores, processes, or handles customer data as a part of what it sells. This could include:
- SaaS and cloud service providers
- Data, analytics, or AI platforms
- Hosting, IT, and infrastructure service providers
- Financial technology, Healthcare, and HR service providers that sell to other companies
- Vendor products that plug into customer systems and data
What it covers: SOC 2 evaluates an organization’s controls based on five operational areas, known as ‘Trust Services Criteria’. These are:
Of these, Security is mandatory, while the others are included based on the nature of services offered.
Types of reports: There are two types of SOC 2 reports—Type 1 and Type 2.
A SOC 2 Type 1 report covers control design at a point in time.
A SOC 2 Type 2 report evaluates if your controls worked as intended over a period of time, typically 3 to 12 months.
Other frameworks and regulations you’ll likely encounter
Think of SOC 2 as your home planet. Everything else in the galaxy is a short trip from here.
It’s a common starting point for most early-stage companies, but it isn’t the only way in. It does, however, become your foundation while the ones listed below apply to you depending on which regions you sell to and what data you handle. You can pursue them alongside SOC 2 or even earlier if your market calls for it.
Unlike SOC 2, ISO 27001 is a certification from an accredited body. It sets out guidelines on building and maintaining what the International Organization for Standardization (ISO) calls an Information Security Management System (ISMS).
An ISMS is a structured set of policies, procedures, and technical controls designed to identify information security risks, evaluate them, and implement measures to minimize impact.
While US buyers prefer SOC 2, customers from Europe, the UK, the Middle East, and most of Asia prefer ISO 27001. Many founders who relocate to the US end up selling across two or more of these regions. So a combination of both frameworks is very common.
A significant number of ISO 27001 controls overlap with SOC 2. So if you’ve got the groundwork done for SOC 2, ISO 27001 becomes an extension of these efforts rather than a new project altogether.
The Health Insurance Portability and Accountability Act, HIPAA for short, is a US federal law and comes into effect if you handle Protected Health Information (PHI) for hospitals, clinics, insurers, or other healthcare organizations, including as their software provider. HIPAA is overseen by the US Department of Health and Human Services (HHS), which issues its rules, and enforced by the HHS Office for Civil Rights (OCR).
HIPAA has three key rules:
- Privacy rule (provides patients with certain rights, affording them better control over their PHI, and limiting how this data is used)
- Security rule (ensures the security of PHI that is collected, stored, processed, transmitted, and maintained by Covered Entities)
- Breach notification rule (requires Covered Entities and Business Associates to notify impacted parties and regulators about breaches within a stipulated timeframe)
It’s important to note that HIPAA does not have an official certification. Failing to comply with HIPAA carries hefty penalties, including fines, terminations, sanctions, and criminal charges.
HIPAA rules apply to two types of organizations—Covered Entities (healthcare providers, health plans, and healthcare clearinghouses), and Business Associates (organizations that handle PHI on behalf of Covered Entities or provide them with services).
If you’re a startup with healthcare organizations as customers, HIPAA applies to you. Customers will expect you to sign a BAA (Business Associate Agreement), a binding contract that guarantees you will keep their information safe.
SOC 2 covers a lot of the groundwork for HIPAA. HIPAA’s Security Rule specifically evaluates administrative, physical, and technical safeguards, and many of these overlap with SOC 2.
Having said that, HIPAA includes unique measures, such as the Breach Notification Rule, which sets deadlines for notifying affected parties and regulators.
The Payment Card Industry Data Security Standard, PCI DSS, for short, is a global security standard for the protection of cardholder data. First released in 2004, PCI DSS has been managed by the PCI Security Standards Council, founded by American Express, JCB, Discover, Visa, and Mastercard.
PCI DSS essentially covers security measures designed to ensure the safe and secure handling of cardholder data. The framework groups 12 core requirements under 6 vital goals:
- Build and maintain a secure network and systems
- Protect account data
- Maintain a vulnerability management program
- Implement strong access control measures
- Regularly monitor and test networks
- Maintain an Information Security Policy
Unlike SOC 2 and ISO 27001, non-compliance incurs fines ranging from $5k to $100k per month, depending on the duration. A breach can add to these costs, and companies can even lose the ability to accept card payments.
PCI DSS applies to any organization that collects, processes, transmits, and stores cardholder information. PCI DSS groups organizations into various levels based on the number of transactions processed per year.
PCI DSS is validated through a Self-Assessment Questionnaire or by a Qualified Security Assessor, depending on the organization's level.
Many controls from SOC 2 count towards both frameworks—MFA, unique accounts, encryption, written policies, and vendor reviews, to name a few. And while SOC 2 provides a strong foundation, PCI DSS requires its own scoping and validation.
ISO 42001 is the first international standard for AI management. Launched in 2023, it provides organizations with a structured approach to governing the development and use of AI. At its core, ISO 42001 provides guidelines for building and maintaining an AI Management System (AIMS).
Like ISO 27001, it’s a certification issued by an accredited body.
ISO 42001 applies to any organization that builds, uses, or sells AI systems. Startups fine-tuning their models or building features on a third-party model API can pursue the framework. While the certification is voluntary, it’s become a real differentiator for any company with AI at its center, especially for highly regulated industries like healthtech or finance.
ISO 42001 and SOC 2 overlap significantly in the foundational security layer. Risk assessment, policies, defined roles, and access controls are some ways they overlap. What SOC 2 doesn’t cover is the AI-specific layer, which includes measures such as human oversight, transparency, lifecycle controls, and impact assessments.
NIST, or the National Institute of Standards and Technology, is the institute that formalized its AI Risk Management Framework in 2023. It serves as a voluntary standard governing the management of risks arising from the development and use of AI.
The framework is built around four key functions:
- Govern: Setting up policies, roles, and culture for managing AI risk across the AI lifecycle
- Map: Understanding the context of AI—what it does, who it impacts, and what could go wrong
- Measure: Assessing and tracking risks through testing and monitoring
- Manage: Prioritizing risks, determining response, and keeping watch when things change
NIST AI RMF is designed for any organization that builds, deploys, or uses AI, regardless of size and industry. This includes startups building AI features on a third-party model API, not just large companies looking to build their own models.
The framework is voluntary, but matters most for AI-first companies looking to sell to US enterprises or working in regulated industries. Some state AI laws also designate it as a recognized framework.
Similar to ISO 42001, SOC 2 and NIST AI RMF have significant operational overlaps. The Govern step leans on the same written policies and defined roles under SOC 2. The Map and Measure steps build on SOC 2’s risk assessment and control monitoring, while the Manage step mirrors risk management and mitigation.
SOC 2 does not include the AI-specific layer addressing bias, fairness, explainability, and impact assessment.
NIST AI RMF is not a certification and does not involve an audit. That’s why many startups in the US align with the framework before moving on to ISO 42001 when customers request AI certification.
To sum up
See the whole table
You don’t need everything at once
Once you start reading about compliance, it sounds like an unending list of frameworks you need all at once. So here’s an important myth worth killing. Trying to do everything at once is how small teams burn out. Focus on one framework at a time, and add the next when customers ask.