Security
Privacy, security, and compliance posture.
Protocol Wealth is an SEC-registered investment adviser. This page summarizes how we protect client information, supervise AI-assisted workflows, manage vendors, and preserve required records.
Last reviewed September 25, 2026. The linked PDF is the current public posture artifact for website use. The sections on cryptography, on how we build software and on the providers we rely on were added in October 2026 and are not yet reflected in the PDF.
This is a public summary, not an exhaustive control report, SOC 2 report, ISO certificate, penetration-test report, client agreement, or replacement for our Form ADV, Privacy Notice, Terms of Service, or other required disclosures. We state material limitations plainly because public security communications should be fair, balanced, and verifiable.
Core commitments
What clients and reviewers should know first
Client data is not sold or used to train AI models.
Vendor relationships are limited to providing services to the firm and are governed through our Privacy Policy, Subprocessors Inventory, and vendor due-diligence process.
Human advisers remain accountable.
AI assists with research, monitoring, summarization, and drafting. It does not make final investment decisions or send client-facing advice without human adviser review.
Access is restricted, and changes are logged.
Client data access is role-restricted and isolated by database row-level security. Actions that create, change or delete client data are written to an audit log under the firm's books-and-records program.
Required records are retained.
Actions that create, change or delete client data are written to an audit log. Audit records are mirrored to a Google Cloud Storage archive whose retention lock is seven years and cannot be shortened.
Privacy and AI controls
AI is supervised and data-minimized.
Our primary reasoning-model provider is Anthropic's Claude API under a contracted Zero Data Retention configuration. Under that arrangement, API inputs and outputs are not used for training or model improvement, and inference is restricted to US-based infrastructure. Under those terms, Anthropic may keep content its safety systems flag, or that the law requires it to keep, for up to two years. Since September 22, 2026, advisers can also choose xAI's Grok in advisor chat. xAI is engaged under team-level zero data retention, our systems check that assertion on every response (the check runs after the request is sent), and requests go only to xAI's U.S. regional endpoints. Client information in such a conversation can reach xAI.
We also use narrower AI capabilities: adviser-facing image/graphics generation and search indexing of documents added to advisor chat (Google Gemini, through Google Cloud's Vertex AI at U.S. endpoints). The AI assistant in the client portal, a limited beta on Claude, answers a client's own questions directly, without adviser review, and gives no investment advice. Values that AI reads from statements and K-1s you or your adviser provide for outside investments may appear to you before an adviser reviews them; each is labeled "Unverified" until an adviser confirms or corrects it. If you turn on AI Suggestions in the client portal, Claude suggests labels and rules for your own bank transactions without adviser review, and nothing changes until you confirm one. All of these are disclosed in our Privacy Policy and List of Subprocessors.
Client-data fields are classified at ingestion as high, medium, or low sensitivity. Before text is sent to an AI provider, our systems replace the identifiers they recognize with placeholders, such as government identifiers, account numbers, email addresses, phone numbers and dates of birth. Recognition is by pattern, is not complete, and does not remove names. It does not cover PDF and image files attached in advisor chat, which are sent as files. A statement or K-1 you or your adviser provide is different: its text is read from the PDF and scanned in the same way before it is sent. An independent egress canary can block outbound requests when residual sensitive patterns are detected.
Controls
Current security controls
Data protection
- TLS 1.2 or better for external traffic, with TLS 1.3 preferred.
- Cloud SQL and Cloud Storage encrypted at rest with AES-256.
- Application-layer AES-256-GCM protection for sensitive vendor credentials.
- Private-network production Cloud SQL and Redis services.
- Google Workspace Enterprise Plus for firm email, documents, and files, encrypted at rest with AES-256 under Google-managed keys, with Google Vault retention and eDiscovery.
Identity and isolation
- Staff sign in through Google Workspace with two-step verification. Clients sign in with a passkey, Google sign-in, or a one-time link sent to their email address.
- Passkey-first authentication on newer client onboarding flows.
- Database-enforced row-level security for tenant isolation.
- Workload identity for non-human access; long-lived service-account keys disabled.
Monitoring and response
- Cloud-configuration-change alerts for high-risk GCP changes.
- Google Security Command Center findings feed our operational security work.
- Dependency vulnerability scanning, secret scanning, and PII-egress alerts.
- Written incident response plan and post-incident review process.
Records and vendors
- Canonical audit log for state-changing actions.
- Seven-year, append-only WORM archive for audit records. Protocol Wealth's own recordkeeping obligation as an adviser is SEC Rule 204-2.
- Subprocessor review before onboarding and periodically after that.
- Vendor diligence evidence available under NDA where redistribution terms permit.
Cryptography
The cryptography behind our sites and wallets.
What each of our sites negotiated when we checked it on October 20, 2026, and the signature schemes the blockchain networks themselves use. Providers can change these settings over time.
- This website (protocolwealthllc.com)
- When we checked, it was served through Cloudflare over TLS 1.3, with a certificate using an ECDSA key on the NIST P-256 curve. With browsers that support it, the connection key between your browser and Cloudflare is agreed using X25519MLKEM768, a hybrid that combines the classical X25519 exchange with ML-KEM-768, the post-quantum key exchange NIST standardized in FIPS 203. Older browsers use a classical key exchange.
- Client portal and adviser platform (pwportal.app, pwos.app)
- Served through Google Cloud's load balancer over TLS 1.3, with a 2048-bit RSA certificate. In our checks the connection key was agreed using classical X25519. We do not claim post-quantum key exchange for these two sites.
- HTTPS only
- All three sites send a Strict-Transport-Security header, so browsers refuse to load them over plain HTTP. protocolwealthllc.com is on the HSTS preload list built into major browsers, and every .app domain is preloaded by its registry, so this applies even on a first visit.
- Wallet signatures
- Each blockchain sets the signature scheme its transactions must use. Bitcoin and Ethereum-compatible networks use ECDSA or Schnorr signatures on the secp256k1 curve, and Solana uses Ed25519. Neither is one of the NIST P-curves, such as P-256, that the research below concerns. The passkey that secures the self-custodial wallet in our client portal typically signs with ECDSA on P-256, as passkeys generally do.
- Recent research on the NIST curves
- On October 16, 2026, Anthropic published research showing how the seeds of the NIST elliptic curves, including P-256, were generated in 1997 and 1998. In its words: "This is not a cryptographic break, new attack, or weakness in the ECDSA curves." The research does not address secp256k1 or Ed25519.
- Quantum computing
- A large enough quantum computer could break all of the classical public-key methods above: RSA, P-256 (including passkeys), X25519, secp256k1 and Ed25519. This is an industry-wide risk, not one specific to us. The hybrid key exchange on this website protects the secrecy of the connection between your browser and Cloudflare, including against someone who records it now and tries to decrypt it later; it does not change the certificate, passkeys or any wallet signature. Bitcoin, Ethereum and Solana do not offer a post-quantum signature scheme in use today. None of our systems is quantum-proof.
Software development
How we build and review software.
We build our own client portal, adviser platform, and the services behind them. We use AI tools heavily to write and review that software. A firm partner directs that work and makes the design decisions. The AI tools work under the controls below.
Changes go through pull requests.
The code for our client portal, adviser platform, core data service, and cloud infrastructure is kept in version control. A change reaches the main version only through a pull request. The code host enforces this rule, and it applies to administrators too. Every member of our code-host organization must use two-factor authentication, limited to authenticator apps, the GitHub mobile app, passkeys, and security keys (no text-message codes).
Review is sized to the risk.
Our practice is that a separate AI review reads a change before it is merged. Changes that may carry client information are reviewed on AI providers' zero-data-retention terms. Higher-risk changes, such as those touching sign-in, permissions, the audit log, money movement, or client data, normally get a deeper review that includes a second AI vendor. By practice, the session that wrote a change does not review it. This is an engineering practice, not an automated gate, and a person does not read every line of code.
People own regulated wording.
A firm partner reviews marketing copy before it is published. Changes to this website must also pass an automated check for Marketing Rule issues before they can be published. That check supports the partner's review; it does not replace it.
Automated checks run before merge.
Automated checks run on our development machine before code is pushed, and on the code host before it is merged. Depending on the code, they include style and type checks, tests, builds, and infrastructure plans.
Secrets and vulnerabilities are scanned.
Pull requests are scanned for passwords and keys committed by mistake. Software dependencies are checked for known vulnerabilities, and security updates arrive as automatic pull requests. Our core data service and its container image, and our infrastructure configuration, are also scanned for known vulnerabilities and misconfigurations.
Infrastructure is defined in code.
Most of our cloud infrastructure is written as code (Terraform), so its settings are versioned and reviewed like any other change. Changes made through it are planned on their pull request and applied to production by a deliberate manual step after merge, not automatically on merge.
Releases are tied to an exact version and checked.
Each production release of our client portal, adviser platform, and data service is built from the exact version of the code that was merged. Our practice is to check each release after it is deployed, before we report it done.
Providers
Providers and tools we rely on.
These are the main providers and tools behind the controls on this page. Some of these providers handle client information for us; for each one that does, what it receives and where, see our List of Subprocessors.
- Google Cloud
- Hosts our client portal, adviser platform, databases, and file storage in the United States. Databases sit on a private network. Google's Gemini models, on Google Cloud, index documents advisers add to advisor chat. Those passages can include client information and are not under zero-data-retention terms.
- Google Workspace, with Google Vault
- Firm email, documents, calendar, and phone. Google Vault provides retention and eDiscovery for them.
- Cloudflare
- DNS for our domains, and hosting for this public website and its forms. Client portal and adviser platform traffic does not pass through Cloudflare.
- GitHub
- Holds our source code and its change history, and runs the automated checks described above.
- Anthropic (Claude)
- Our primary AI provider. Advisor chat and the client portal assistant (limited beta) use Claude under zero-data-retention terms, and so does AI Suggestions in the client portal if the client turns it on; it receives each of the client's own bank transactions' description, merchant name and bank category, whether it is money in or out, the type of account, and how it is currently grouped. Under those terms, Anthropic may keep content its safety systems flag, or that the law requires it to keep, for up to two years. Claude also helps write and review our software.
- xAI (Grok)
- A second AI provider, used under xAI's zero-data-retention terms for our team. Advisers can choose it in advisor chat, and it helps write and review our software.
- Fordefi
- MPC wallet technology for managed onchain strategies, described below. Fordefi is not a qualified custodian.
- ClamAV
- An open-source virus scanner that runs inside our own systems, so files are not sent to an outside service for scanning. Files clients send through our secure upload page are held in quarantine and reach an adviser only after a clean scan.
Digital assets
Onchain wallets: who holds the keys.
The self-custodial wallet in the client portal, built on Turnkey, is being offered first to selected clients. Clients it is offered to can turn it on in the client portal. The client is its sole owner, and it is secured by the client’s passkey. Protocol Wealth never holds the private key, can never sign on the client’s behalf, and does not recover wallets, so the portal asks each client to add a backup passkey on another device.
For active onchain strategies, Protocol Wealth may also operate a whitelist-bounded managed sleeve on Fordefi MPC wallet technology. The client administers the wallet's Rules Policy and holds final approval; Protocol Wealth can transact only to client-pre-approved addresses and contracts (which may include third-party DeFi contracts) and cannot add destinations or move assets outside that whitelist. Signing authority is divided into cryptographic shards: Fordefi holds one shard and Protocol Wealth holds one shard. The client holds the ultimate backup, and Protocol Wealth cannot recover the wallet on its own. Fordefi is wallet technology, not a qualified custodian.
Compliance framework
Regulated adviser obligations come first.
Protocol Wealth's security program is designed around its obligations as an SEC-registered investment adviser, including Regulation S-P safeguards and breach-notification readiness, Rule 204-2 books and records, fiduciary supervision, and Marketing Rule review for public communications.
Our Regulation S-P posture is to maintain reasonable measures to safeguard client information and to provide any legally required breach notification without undue delay, within the deadlines set by applicable law and regulation. We do not represent a fixed contractual 72-hour notification clause; internal response targets do not override the deadlines the rule actually sets.
The firm maintains an ISO/IEC 27001:2022-aligned Information Security Management System with a Statement of Applicability and Risk Register. This is an internal control framework and diligence artifact. Protocol Wealth is not ISO 27001 certified.
Protocol Wealth does not currently hold a firm-level SOC 2 report. We rely on vendor SOC 2 reports where appropriate and maintain a readiness path in case a future customer, partner, or commercial need justifies a firm-held SOC 2 engagement.
Where FINRA standards are relevant through broker-dealer partners or public-communications discipline, we draft communications to be fair, balanced, and not misleading. Protocol Wealth does not present this page as a FINRA member communication unless separately approved through the appropriate channel.
Current limitations
What is not being claimed.
- Customer-managed encryption keys (CMEK) and column-level field encryption are not in place.
- No formal external penetration test has been completed.
- No scheduled access-review cadence is in operation.
- The independent PII egress check does not yet cover every API egress path.
- Our PII scanning does not remove names, and PDF or image attachments in advisor chat are not scanned.
- We have not yet obtained xAI's SOC 2 report for our vendor file.
- ISO 27001 certification and a firm-level SOC 2 report are not current claims.
- Review of software changes by AI is an engineering practice, not an automated gate, and a person does not read every line of code.
- Virus scanning covers files sent through our secure upload page, not every way a file can reach us.
- Of the three sites described under Cryptography, post-quantum key exchange is in use only on protocolwealthllc.com. The client portal, the adviser platform, passkeys and blockchain wallet signatures use classical public-key cryptography.
- No set of controls makes a system impossible to breach. These controls reduce risk; they do not eliminate it.
Verify
Diligence materials available under NDA.
Qualified partners and institutional reviewers may request supporting diligence materials, including vendor due-diligence evidence, anonymized audit-log samples, the ISO-aligned Statement of Applicability, the Risk Register, and security-policy artifacts.
Attestation
Current posture artifact.
This public security posture was last reviewed on September 25, 2026. The linked PDF is the current public posture artifact for website use.
Download the public posture PDFRead next
Related pages
Custody and Control
How client assets are held, who controls them, and why Protocol Wealth is not a qualified custodian.
Self-Custody & Wallets
The onchain wallet arrangements described above, in more detail.
Trust Center
The firm's full set of trust surfaces — privacy, subprocessors, vendors, AI, and disclosures.
Last updated: September 25, 2026. Protocol Wealth LLC is an SEC-registered investment adviser (CRD #335298). Full regulatory disclosures are available through our Form ADV.
Registration with the SEC does not imply a certain level of skill or training. This page is provided for informational purposes and does not create additional contractual obligations beyond those set out in our advisory agreements, Privacy Policy, and Terms of Service.
Legally required notifications, including breach notifications under Regulation S-P, follow the deadlines set by applicable law and regulation. Internal response targets do not override those requirements.
