Information Security Addendum

This Information Security Addendum ("Security Addendum") forms part of the Agreement between Taint Labs and Customer. It also serves as Annex II (Technical and Organizational Measures) to the Data Processing Addendum.

Version
1.0
Published
November 15, 2026
Entity
Taint Labs Corporation

01Scope, Definitions and Precedence

1.1 Scope. This Security Addendum describes the administrative, technical and physical measures Taint Labs maintains to protect Customer Data processed by the Services. It applies for as long as Taint Labs processes Customer Data under the Agreement.

1.2 Definitions. Capitalized terms not defined here have the meanings given in the Master Subscription Agreement or the Data Processing Addendum. In addition:

  • "Test Credentials" means any API key, access token, password, OAuth grant, service account, certificate or other secret that Customer provides to, or makes reachable by, the Services or a Customer Agent for the purpose of an Evaluation Run (the same meaning as in Product Appendix 1 Section 2.7).
  • "Evaluation Environment" means the compute environment in which Taint Labs executes an Evaluation Run, including the execution of attack tests from the Attack Catalog and their benign twin tasks.
  • "Security Incident" means a breach of security leading to the accidental or unlawful destruction, loss, alteration, unauthorized disclosure of, or access to, Customer Data in Taint Labs' possession or control. Unsuccessful attempts that do not compromise Customer Data (for example, blocked scans, rate-limited requests, or failed sign-in attempts) are not Security Incidents.
  • "Production Systems" means the systems Taint Labs uses to provide the Services to Customer.

1.3 Order of precedence. If documents forming the Agreement conflict, the following order applies, highest first: (1) the Order Form, only for commercial terms it expressly overrides; (2) the Data Processing Addendum, for processing of Personal Data only; (3) this Security Addendum; (4) the Product Appendix; (5) the Master Subscription Agreement; (6) the Support and Service Level Policy and the Acceptable Use Policy. Early Access Terms control over all of the above for Early Access Services only.

1.4 Early Access Services. Stryct Finder is currently offered as an Early Access Service.

1.5 No certification. [●] Taint Labs does not hold, and does not represent that it holds, any security certification or third-party attestation, including SOC 2, ISO/IEC 27001, ISO/IEC 27701, ISO/IEC 42001, PCI DSS, HITRUST, FedRAMP or HIPAA-related attestations. Taint Labs has no third-party penetration test attestation to provide. This matches what Taint Labs states publicly in its Privacy Policy (https://stryct.ai/privacy.html, section 12). Nothing in this Security Addendum, the Documentation, or any Results is a certification.

1.6 Measures may improve. Taint Labs may update the measures in this Security Addendum from time to time, provided that no update materially reduces the overall level of protection for Customer Data during the Subscription Term.

02Security Program and Governance

2.1 Program. Taint Labs maintains an information security program appropriate to its size, the nature of the Services, and the sensitivity of Customer Data. [●]

2.2 Responsibility. A named individual is responsible for the security program. [●]

2.3 Review. Taint Labs reviews the program at least [●] and after any Security Incident or material change to the Services.

2.4 Risk assessment. [●]

03Access Control

3.1 Least privilege. Access to Customer Data and Production Systems is limited to Taint Labs personnel who need it to provide, support or secure the Services, and only to the extent they need it. For the public website, Taint Labs states publicly that access to its waitlist database, analytics database and vulnerability reports is limited to the small number of people who need it for their work (Privacy Policy, section 12). [●]

3.2 Authentication for administrative access. Administrative access to Production Systems and to the accounts of providers that host them (including the Cloudflare account, the source-code host, and the email provider) requires multi-factor authentication. [●]

3.3 Unique accounts. Each person with access uses an individual account. Shared administrative credentials are not used. [●]

3.4 Access review and offboarding. Access is reviewed at least [●] and removed promptly, and in any case within [●], when a person no longer needs it or leaves Taint Labs. [●]

3.5 Customer User access. Users sign in to Taint Labs' account pages through Taint Labs' authentication provider (Clerk). [●]

3.6 Service-to-service authentication. Inbound webhooks from Taint Labs' authentication provider are accepted only when their HMAC-SHA256 signature verifies against a secret held by Taint Labs. [●]

04Encryption

4.1 In transit. Traffic to stryct.ai and docs.stryct.ai is served over TLS through Cloudflare (Privacy Policy, section 12). [●]

4.2 At rest. Customer Data at rest is encrypted. [●]

4.3 Secrets. Application secrets for the stryct.ai website and its intake endpoints (signing keys for form tokens, analytics and webhooks; email provider API keys) are held as environment secrets at the hosting provider and are not stored in source code. [●]

4.4 Hashing and tokens. Where the stryct.ai website needs to recognize a value without keeping it, it stores a keyed or cryptographic hash rather than the value (for example, the acknowledgment rate-limit counter stores a SHA-256 hash of the recipient address, not the address; Privacy Policy section 12). Form-submission and analytics tokens are signed with HMAC-SHA256 and compared in constant time.

05Logical Separation of Customer Data

5.1 Separation. Customer Data is logically separated from the data of other customers, so that one customer's Users cannot access another customer's Customer Data or Results. [●]

5.2 Separation from public data. Customer Results are never published on Taint Labs' public leaderboard without Customer's express written opt-in. The public leaderboard is compiled ahead of time from evaluation reports that Taint Labs itself selects and holds, and is not populated from Customer accounts. [●]

5.3 Separation from website data. The website's waitlist and analytics data are stored in two separate databases. [●]

06Secure Development and Change Management

6.1 Source control. Taint Labs maintains its source code in version control, and changes are recorded as commits. [●]

6.2 Automated checks. Taint Labs maintains automated build validation and test suites for the stryct.ai website and its intake endpoints. [●]

6.3 Dependency management. [●]

6.4 Secure coding practices. Where the stryct.ai website and its intake endpoints handle untrusted input they apply defensive measures, including: escaping all submitted values before they are placed in HTML email; rejecting cross-origin form posts; bounding request body size to 64 KB; using parameterized database queries; and setting response headers that prevent content-type sniffing and caching on intake responses. [●]

6.5 Separation of environments. [●]

6.6 Deployment. [●]

07Vulnerability Management and Coordinated Disclosure

7.1 Public disclosure channel. Taint Labs operates a public coordinated vulnerability disclosure program. Reports go to security@taint.ai or through the form at https://stryct.ai/report.html. The machine-readable contact file at https://stryct.ai/.well-known/security.txt names that page as its policy. The policy includes a safe harbor whose scope is stated in the website Terms of Service, section 09.

7.2 Response target. Taint Labs aims to have a person acknowledge each report within three U.S. business days (https://stryct.ai/report.html, section 06). This is a target, not a guarantee.

7.3 Intake protection. The report intake rejects abusive traffic before it spends storage, using origin checks, content-type checks, per-IP and global rate limits, a bounded body read, a honeypot field, and a signed timing token.

7.4 Internal vulnerability management. [●]

7.5 Findings about Customer Agents. Vulnerabilities that an Evaluation Run finds in a Customer Agent are Results and Customer Data. They are delivered to Customer and are not disclosed to any third party, including the vendor of a model or SDK the Customer Agent uses, except as the Agreement permits or Customer directs in writing.

08Test Credentials Supplied for Evaluation Runs

8.1 Purpose limitation. Taint Labs uses Test Credentials only to perform the Evaluation Runs Customer requests, and for no other purpose.

8.2 Storage. Test Credentials are stored encrypted, separately from Results, and are not written to logs, traces, Results, or support tickets. [●]

8.3 Redaction. [●]

8.4 Retention. Test Credentials are deleted [●].

8.5 Customer revocation. Customer may revoke or rotate any Test Credential at any time. Taint Labs recommends that Customer rotate Test Credentials after each engagement.

09Isolation of Evaluation Runs

9.1 Nature of Evaluation Runs. Evaluation Runs deliberately exercise adversarial inputs, including prompt injection, tool misuse and exfiltration attempts, against a Customer Agent. Taint Labs designs the Evaluation Environment so that this activity is contained.

9.2 Isolation. Each Evaluation Run executes in an isolated environment that is not shared with another customer's Evaluation Runs and is torn down after the run. [●]

9.3 Network egress. [●]

9.4 Synthetic data. Attack payloads and canary values used in Evaluation Runs are synthetic and generated by Taint Labs. [●] Taint Labs does not intentionally cause a Customer Agent to send real Customer Data to a destination outside Customer's or Taint Labs' control. [●]

9.5 Customer-hosted execution. Stryct Finder can be deployed on-premises, with a Customer-Run Runner operating in Customer's environment (Product Appendix 1, Section 4.3). In that mode this Section 9 applies only to components Taint Labs operates. [●]

10Logging and Monitoring

10.1 Logging. The stryct.ai website runs on Cloudflare with Cloudflare's observability logging enabled, and logs operational events, including failures of its scheduled data deletion jobs. [●]

10.2 Log content. Logs are designed not to contain Test Credentials or the content of Customer Data. [●]

10.3 Monitoring and alerting. [●]

10.4 Abuse controls. Public endpoints on the stryct.ai website are protected by Cloudflare rate limits applied before any database read or write, and inherit Cloudflare's network-level DDoS mitigation.

11Security Incident Response and Notification

11.1 Response. Taint Labs maintains a process to identify, contain, investigate, remediate and document Security Incidents. [●]

11.2 Notification to Customer. Taint Labs will notify Customer without undue delay, and in any event within [●], after becoming aware of a Security Incident affecting Customer's Customer Data. Notice goes to the security contact named on the Order Form, or if none, to Customer's account administrator.

11.3 Content of notice. To the extent known, the notice will describe the nature of the Security Incident, the categories of Customer Data affected, the likely consequences, the measures taken or proposed, and a contact point. Taint Labs will provide further information as it becomes available.

11.4 Cooperation. Taint Labs will reasonably cooperate with Customer so that Customer can meet its own legal notification obligations. Taint Labs will not name Customer in any public statement or regulatory notice about the Security Incident without Customer's consent, except as required by law.

11.5 Not an admission. Notification of a Security Incident is not an acknowledgment of fault or liability.

11.6 Consistency with the Privacy Policy. Taint Labs' public Privacy Policy commits to notify the relevant supervisory authority within 72 hours where the law requires it (Privacy Policy, section 12). The DPA governs regulatory notification for Customer Personal Data.

12Business Continuity and Backup

12.1 Backups. [●]

12.2 Continuity. [●]

12.3 Provider resilience. The Services depend on Third-Party Services and hosting providers whose availability Taint Labs does not control. The MSA and the Support and Service Level Policy govern the consequences of downtime.

13Personnel Security

13.1 Confidentiality. All Taint Labs personnel and contractors with access to Customer Data are bound by written confidentiality obligations. [●]

13.2 Background checks. [●]

13.3 Training. Personnel with access to Customer Data receive security and privacy awareness training at onboarding and at least [●]. [●]

13.4 Devices. [●]

14Subprocessor and Vendor Security

14.1 Subprocessors. Taint Labs engages Subprocessors only as permitted by the DPA, and lists them in the Subprocessor List. For the public website, Taint Labs' processors are Cloudflare (hosting, databases, network), Clerk (account authentication) and Resend (transactional email) (Privacy Policy). [●] Any hosted model provider that processes Customer Data under Taint Labs' account is a Subprocessor, is listed on the Subprocessor List before it does so, and is used only on API terms that prohibit training on that data and, where the provider makes them available, on zero- or limited-retention terms (MSA Section 7.4).

14.2 Vendor diligence. Before engaging a Subprocessor that will process Customer Data, Taint Labs reviews its security practices and binds it by written terms to protections no less protective than those in the DPA, to the extent applicable to the service provided. [●]

14.3 Third-Party Services chosen by Customer. Models, APIs, MCP servers and other Third-Party Services that a Customer Agent calls during an Evaluation Run are chosen by Customer and are not Taint Labs' Subprocessors. Taint Labs is not responsible for their security.

15Data Retention, Return and Deletion

15.1 Retention. Taint Labs retains Customer Data only as long as needed to provide the Services and as the Agreement permits. [●]

15.2 Deletion after termination. Taint Labs deletes Customer Data from Production Systems within 30 days after the end of the 30-day Export Window in MSA Section 11.5, or earlier on Customer's written instruction. Copies in backups are deleted within [●] days after that deletion and are not restored to active use in the meantime. Where Customer operates Taint Labs software in its own environment, this Section applies to the Customer Data Taint Labs itself holds (for example, Results or telemetry sent to Taint Labs, support uploads and account data), and Customer Data stored in Customer's environment remains with Customer.

15.3 Return. During the Export Window, Customer has read-only access to export Customer Data, including Results, in [●].

15.4 Deletion is automated where stated. Where Taint Labs publishes a retention period for the stryct.ai website, it is enforced by a scheduled job rather than by hand: a daily job deletes rate-limit counters after 1 day, acknowledgment counters after 30 days, and analytics events and daily rollups after 30 days. Failures are logged. [●]

15.5 Exceptions. Taint Labs may retain Customer Data where required by law, and may retain Aggregated Data only as, and subject to the limits in, MSA Section 7.5. Taint Labs does not use Customer Data to train, fine-tune or otherwise improve any machine-learning model (MSA Section 7.4).

15.6 Certification of deletion. On written request, Taint Labs will confirm in writing that deletion under Section 15.2 has occurred.

16Customer Security Responsibilities

16.1 Shared responsibility. Security of the Services is shared. Taint Labs is responsible for the measures in this Security Addendum. Customer is responsible for the matters in this Section 16 and for its own systems.

16.2 Test credentials. Customer will supply Test Credentials that are dedicated to evaluation, scoped to the minimum permissions an Evaluation Run needs, revocable, and not shared with production workloads. Customer will not supply credentials with access to production data, payment systems, or privileged administrative functions unless a Testing Authorization and Rules of Engagement signed by both parties names the system they reach (MSA Section 5.4).

16.3 Non-production environments. Customer will point Evaluation Runs at non-production or staging instances of Customer Agents, and at test data, wherever possible. Evaluation Runs attempt to induce harmful behavior on purpose; if Customer points them at a production system, Customer accepts the risk that the Customer Agent takes real actions (for example sending a message, modifying a record, or calling a paid API).

16.4 Customer Agent behavior. Customer is responsible for what its Customer Agents do, including actions they take, data they send to Third-Party Services, and costs they incur during an Evaluation Run. Results describe observed behavior under test; they are measurements, not a certification that a Customer Agent is secure.

16.5 Personal Data in test inputs. Customer will not include real Personal Data, or sensitive or regulated data, in Customer Agent configurations, prompts, tool fixtures or test data unless the DPA covers it and Customer has a lawful basis.

16.6 Accounts. Customer is responsible for managing its Users, keeping their sign-in credentials confidential, enabling any available MFA, removing Users who no longer need access, and notifying Taint Labs promptly at security@taint.ai of any suspected compromise of a User account or Test Credential.

16.7 Authority. Customer will only submit or connect Customer Agents and Third-Party Services that Customer owns or is authorized to test, consistent with the Acceptable Use Policy and the Third-Party Services' own terms.

17Security Assessments

17.1 Questionnaire. Once per contract year, on written request, Taint Labs will complete one reasonable security questionnaire (or a standard industry questionnaire Taint Labs chooses to maintain). Taint Labs will answer accurately, including stating plainly where a control does not exist.

17.2 Penetration testing. [●]

17.3 Audit reports. [●] If Taint Labs obtains a third-party audit report in the future, it will make the report available on request under confidentiality, and that report will satisfy Customer's audit rights to the extent it covers them.

17.4 Customer audits. Any additional audit or on-site inspection is governed by the DPA, must be on reasonable notice, during business hours, not more than once per year (except after a Security Incident or where a regulator requires), at Customer's cost, and subject to confidentiality.

17.5 Testing the Services. Customer may not perform penetration testing, scanning or attack simulations against Taint Labs' Production Systems except under the public disclosure policy at https://stryct.ai/report.html or with Taint Labs' prior written consent. Pointing Stryct Finder at Customer's own Customer Agents is not testing of Taint Labs' systems.

18Contact

Security questions and Security Incident reports: security@taint.ai. Privacy questions: privacy@taint.ai. Contract notices: legal@taint.ai.