Implementation Note
This document is for separate acceptance when enabling AI auto-remediation, self-healing, auto-deployment, automated actions or high-permission AI features.
Scope of AI Features
AI features provided by PocketBay may include one-click AI deployment, automatic reading of deployment protocols, packaging, building, auto-containerization, Dockerfile generation, health checks, log analysis, error explanations, remediation suggestions, automatic retries, rollbacks, process self-healing, configuration suggestions, code suggestions, prompt examples, task generation, comment-based review and other automation capabilities.
AI features may be implemented through PocketBay-developed systems, third-party models, third-party APIs, open-source models, rule engines, automation scripts and upstream services. AI functionality may vary by plan, territory, project type, risk level, resource status and upstream availability.
AI features are provided as-is and do not guarantee accuracy, completeness, safety, legality, functionality, production suitability, no extra costs, no incorrect modifications or successful issue resolution.
Authorization and Automated Actions
When you enable AI automation, auto-remediation, auto-deployment, self-healing, automatic retries, automatic rollback, automatic resource adjustment or similar features, you authorize PocketBay, within the permissions you configure, to read logs, errors, project structure, dependencies, configurations, build outputs, runtime status and related context, and to trigger builds, restarts, rollbacks, configuration changes, file generation, environment-variable suggestions, API calls, repository connections or upstream services.
You should assess risks and configure least privilege before enabling automation. You are responsible for actions, results, costs, data changes, service interruptions and end-user impact caused by AI acting under your authorization, account, team permissions, API keys, secrets, third-party authorizations or configurations.
PocketBay may suspend, restrict, withdraw or modify AI automation capabilities for security, compliance, cost, abuse, abnormal result or platform stability reasons.
Routine run-status checks, log reading, configured retries or restarts may be performed within the scope you have expressly authorized. Deleting business data, changing access permissions, publishing private content, payments or refunds, sending external messages, changing merchant settlement settings, or spending beyond an authorized budget are high-impact operations requiring the corresponding independent authorization and, where necessary, human confirmation.
Public review uses non-destructive tests and appropriate test data by default and does not trigger real payments, refunds, production messages or irreversible business changes. Free review quotas are displayed explicitly; overage charges are not presumed merely from "AI automation being enabled". The platform applies reasonable controls on task frequency, loops and failure retries.
Output Risks and Human Review
AI outputs may contain erroneous commands, unsafe configurations, outdated dependencies, incorrect explanations, false positives, false negatives, incomplete fixes, license-violating suggestions, performance issues, security vulnerabilities, privacy risks, compliance risks or content unsuitable for your business. You must not rely on AI outputs as legal, medical, financial, tax, security, compliance or professional advice.
For production environments, paid services, Discover listings, personal-information processing, sensitive-data processing or matters affecting end-user rights, you should perform human review, testing, backups, rollback planning, security scanning, dependency review, log redaction and compliance checks.
You should review important AI outputs according to your use; the platform is simultaneously responsible for its own permission design, task execution, security controls, promised services and legally required obligations. The platform does not treat every action taken in platform AI features as your own act, regardless of permissions, fault or causation, or make you bear all responsibility.
Where generated-content labeling rules apply, the platform performs the explicit, implicit or other statutory labeling duties corresponding to its role as generator or disseminator. An ordinary "AI" icon does not replace the complete labeling the law requires for specific scenarios, and the platform does not assist in removing labels that must be retained by law.
AI Data Processing and Third-Party Models
To provide AI features, PocketBay may process project metadata, logs, error stacks, configurations, file paths, build outputs, code snippets, dependency information, deployment status, user instructions and other context. You should avoid submitting unnecessary secrets, personal information, sensitive data, trade secrets or regulated data to AI features.
PocketBay may use third-party AI models or APIs. Third-party model availability, performance, pricing, territories, content policies, security, data processing and output quality may change. PocketBay may replace, restrict, discontinue or add model providers without maintaining any specific model.
You are responsible for ensuring you have the right to use relevant data for AI analysis, auto-remediation and third-party model processing, and for complying with applicable law, privacy policies, customer agreements and third-party terms.
Bayo, Memory and Automated Review
Bayo and related AI features access only the designated project materials, within valid authorization and current permissions. Cross-project retrieval, public display and copy use each check the corresponding permissions; being able to issue AI requests from one account does not cross organization, user or project isolation. Generating public Genome does not authorize AI to publish full Memory, and public retrieval does not trace back to unauthorized private fragments. Model and tool access follows the Project Memory and Genome Data Policy and the platform's actual disclosures.
For Public projects with the feature enabled, the platform may conduct automated functional reviews within the test scope disclosed on the deployment page and display non-sensitive summaries on the project page. Such results are prominently labeled as produced by AI or automated testing and never impersonate real consumer reviews, purchase records or independent human certification.
A review reflects observations at a specific time, version, test scope and environment; it is not a comprehensive security audit, compliance certification, fitness guarantee or traffic or revenue commitment. Material conclusions come with verifiable test conditions and evidence explanations where possible, together with a correction or appeal channel. Review logs, vulnerability details, access tokens, private Memory, real customer information and other risk materials are not disclosed with public summaries. Private, Protected or authenticated real business pages gain no operating permissions by default from routine public review; legally necessary security checks follow security rules at the minimum scope.