Agency_

🛡️Defense-in-Depth Architecture

Security & Trust Center

How Agency protects your workspace, encrypted credentials, connected accounts, and autonomous AI workflows.

Author: Agency•Last Reviewed: November 2026
🔐

AES-256 & TLS 1.3

All traffic is encrypted in transit via TLS 1.3 with HSTS, and connected OAuth/API credentials are encrypted at rest using authenticated AES-256-GCM.

🏢

Strict Tenant Isolation

Every project, document briefcase, email thread, and workflow is cryptographically scoped to your authenticated account with Row-Level Security.

🧠

Zero AI Model Training

We use paid commercial enterprise AI API tiers. Your prompts, documents, emails, and business data are never used to train foundation models.

⚖️

Human-in-the-Loop Gates

External actions—such as sending emails, publishing articles, or posting to social networks—support explicit approval gates and Prime Directive checks.

1. Defense-in-Depth Security Philosophy

At thisisagency.ai, security is engineered into every layer of the platform rather than bolted on as an afterthought. Because an agentic workspace coordinates across your websites, documents, analytics, and communication channels, we enforce a multi-layered Defense-in-Depth architecture that validates identity, permissions, data boundaries, and AI safety at every step of execution.

No single control is trusted in isolation: edge network protections, cryptographic session verification, database row-level isolation, outbound network guards, and autonomous compliance agents operate together so that a failure or abuse attempt at one boundary is contained by the next.

2. Encryption in Transit & Credential Vault at Rest

Encryption in Transit (TLS 1.3 & HSTS)

All browser, API, webhook, and real-time voice connections to Agency are encrypted over HTTPS using TLS 1.2/1.3 with HTTP Strict Transport Security (HSTS) preloading enabled. Unencrypted HTTP traffic is automatically upgraded or rejected at the network edge.

Authenticated Encryption at Rest (AES-256-GCM)

Database storage and automated backups are encrypted at rest on EU-hosted infrastructure. Third-party OAuth tokens, refresh tokens, and integration keys you connect to your workspace are additionally sealed inside an application-level credential vault using authenticated AES-256-GCM encryption with unique initialization vectors prior to database persistence.

3. Identity, Authentication & Session Security

  • Password & Credential Hardening: User passwords are never stored in plaintext; they are salted and hashed using industry-standard adaptive cryptographic hashing (`bcrypt`). Sign-in with Google OAuth 2.0 is supported for passwordless authentication.
  • Cryptographically Signed Sessions: Authenticated sessions use `HttpOnly`, `Secure`, `SameSite` cookies signed with HMAC-SHA256 and verified at the edge on every request, preventing client-side script access or token tampering.
  • Unrecognised Device Alerts & Instant Session Revocation: Sign-ins from new devices or browsers trigger automatic security notifications, and password resets or security events immediately invalidate prior sessions across all devices.
  • Brute-Force & Automated Abuse Protection: Authentication endpoints are protected by Cloudflare Turnstile challenges, progressive rate limiting, and automated IP lockouts against credential-stuffing attacks.

4. Multi-Tenant Workspace & Database Isolation

Agency is architected to ensure strict logical separation between customer accounts and projects:

Row-Level Security (RLS) & Scoped Queries

All production tables enforce PostgreSQL Row-Level Security (RLS), and direct public data API exposure is locked down. Every query for projects, reports, Briefcase documents, financial ledgers, and chat histories is explicitly scoped to the verified session owner.

Scoped Agent Memory & Briefcase Isolation

When our AI specialists read from your Project Briefcase, CEO Knowledge Panel, or connected analytics, context assembly is strictly isolated to your active project. One customer's agents can never query, read, or reference another customer's documents or memory.

5. AI Safety, Zero Model Training & Autonomous Governance

Deploying autonomous multi-agent workflows requires dedicated safeguards against prompt injection, hallucination, and unauthorized external actions:

Zero Foundation Model Training & Google API Limited Use

We access foundation models (including Google Gemini, Anthropic Claude, and OpenAI) strictly through commercial API agreements under which customer inputs and outputs are not used to train or improve public foundation models. Our integration with Google Workspace, Gmail, Calendar, Search Console, and Analytics complies fully with the Google API Services User Data Policy, including its Limited Use requirements.

Prompt-Injection & Jailbreak Defense

Untrusted external content (such as scraped web pages, inbound emails, or caller transcripts) is sanitized and isolated before entering agent context windows. Dedicated security orchestration monitors interactions for prompt-injection and system-extraction attempts.

Prime Directives & Quality Control Gates

Every automated workflow is governed by mechanical Prime Directives (Do No Harm and Ground Truth Above All) and multi-agent Quality Control. Outbound communications and publishing actions can be configured to require explicit Human-in-the-Loop approval before dispatch.

6. Network Edge, Web Application & Billing Security

  • Edge WAF & Distributed Rate Limiting: Every request passes through edge middleware enforcing strict Content Security Policy (CSP), Cross-Site Request Forgery (CSRF) origin verification, automated scanner blocking, and distributed per-IP and per-account rate limiting.
  • XSS & Output Sanitization: Rich HTML and Markdown outputs rendered across the platform pass through strict DOMPurify sanitization and content-type validation to prevent Cross-Site Scripting (XSS).
  • Outbound Crawler & Webhook Hardening: When Agency crawls a public webpage you submit for SEO or quality analysis, outbound requests are validated against Server-Side Request Forgery (SSRF) and DNS-rebinding protections to ensure only public web destinations can be fetched.
  • PCI-DSS Compliant Payments via Stripe: All credit purchases and subscription billing are processed directly by Stripe. Agency never receives, processes, or stores your raw payment card numbers or CVV codes.

7. UK GDPR Compliance, Data Residency & Your Controls

Agency is operated by Candlestickoperations Ltd (registered in England & Wales, Company No. 17503419) in compliance with the UK General Data Protection Regulation (UK GDPR) and the Data Protection Act 2018.

  • EU Primary Database Residency: Primary application databases are hosted in the European Union (`eu-central-1`, Frankfurt) with encrypted automated backups.
  • Instant OAuth Revocation: You can disconnect any third-party integration (Google, Microsoft Bing, CMS, or social accounts) at any time from your Project or Settings dashboard, immediately deleting stored tokens from our vault.
  • Right to Erasure & Export: You may delete individual documents, threads, or projects at any time, or request complete account deletion and data export via our Privacy Policy.

8. Vulnerability Reporting & Security Contact

We welcome collaboration with security researchers and enterprise security teams. If you believe you have discovered a security vulnerability in thisisagency.ai, or if your organization requires a security questionnaire or Data Processing Agreement (DPA), please contact us directly:

Agency (Thisisagency.ai) is in public beta. Agency is an AI workplace simulation. Agency was conceived, bootstrapped, and architected solo by Shaun Anderson AKA Hobo (@hobo_web). “This Is Agency” and “Agency” are trading names of Candlestickoperations Ltd. Registered in England and Wales, Company No. 17503419. Registered Office: 3rd Floor, 86-90 Paul Street, London, England, EC2A 4NE.
All team members are AI entities powered by Google Gemini technology (by default). When API are available, Agency is a multi-provider, multi-model, multi-modal, multi agent system.