# Faradex — Full Content > This file contains the full text of Faradex's public marketing pages, concatenated for AI/LLM ingestion. For a shorter overview with links, see [llms.txt](https://faradex.ai/llms.txt). For the live site, see [faradex.ai](https://faradex.ai/). Faradex is built by **Faraday Collective Inc.** Each section below corresponds to one public page. URLs are production. --- # Landing — `https://faradex.ai/` ## The Power of Agentic AI. Built for Legal. Built for Accounting. Built for Finance. A secure workforce of AI executives, working alongside your team. 24/7. ### What Faradex is - **AI Workforce.** Move beyond ChatGPT and Copilot. Faradex gives you a team of expert AI analysts that work on their own — drafting reports, analyzing data, and handling research around the clock. - **Team Collaboration.** Bring your team and your AI executives into the same room. Faradex lets you strategize together with multiple AI specialists at once, and even loop in clients or outside parties. - **Security First.** Unparalleled security and data privacy. Get the power of the world's best AI without sending sensitive information to OpenAI or Anthropic. ### Get Started in Minutes — three steps 1. **Create Your Faradex.** Spin up your dedicated environment in minutes. Upload your sensitive company documents to a secure vault. PDFs, Word docs, spreadsheets — we handle them all. 2. **Invite Your Team.** Only you and the people you invite can access your instance. Your files, conversations, and AI outputs are inaccessible to everyone else, including us. 3. **Put Your Agents to Work.** Your team interacts with each other and with AI agents who can research, draft, analyze, and extract. They work in parallel, even while you sleep. ### Faradex vs. The Field | Feature | Vertical AI (Harvey, CoCounsel) | Consumer AI (ChatGPT, Copilot, Claude) | Agentic Tools (Cowork, Claude Code) | Faradex | |---|---|---|---|---| | Agentic AI | ✗ | ✗ | ✓ | ✓ | | Private Server | ✗ | ✗ | ✗ | ✓ | | Team Collaboration | ✗ | ✗ | ✗ | ✓ | | No public disclosure risk | ✓ | ? | ? | ✓ | | Zero document retention | ✗ | ✗ | ✗ | ✓ | | Only you can access your files | ✗ | ✗ | ✗ | ✓ | | AI providers never see your data | ✓ | ✗ | ✗ | ✓ | ### Unparalleled Security A new era demanded a new approach to security. Faradex was designed from the ground up for the way firms will work with powerful agentic AIs. Your files and conversations never touch a hard drive. They live in temporary memory inside a server that belongs only to your firm. No one else can see in. Not other customers, not Faradex, not Amazon. Every 24 hours, the entire environment is destroyed and rebuilt clean. The AI models that power Faradex are covered by Amazon's Zero Data Retention. Nothing is stored, so your conversations and files could not be provided even under subpoena. **You cannot subpoena what does not exist. You cannot preserve what was never stored.** ### Trust signals SOC 2 Type I (Type II Q3 2026) · GDPR Ready · End-to-End Encrypted · Zero Data Retention · Single-Tenant Infrastructure ### Closing line **Privacy should not be the cost of using AI.** --- # Security — `https://faradex.ai/security` ## World-class agentic AI. Zero retention. Zero exposure. Zero compromise. Agentic AI tools like Cowork and Claude Code get their power by reaching into your local machine: your filesystem, your browser, your desktop. That's how they build context. But for professionals handling sensitive client data in legal, finance, and medical fields, that's a risk you can't take. > Faradex uses a nothing-at-rest architecture. No database to breach. No files to steal. No backups to subpoena. ### The Faradex Security Boundary Your documents, conversations, and AI agents all stay inside a boundary your firm controls. Zero training. Zero file retention. Isolated infrastructure. ### How your data is secured — four pillars **1. Your own server, your keys only.** Every customer gets a dedicated, single-tenant instance. Your own compute, network, and encryption. No infrastructure is shared with other customers. Only people you invite can get in. **2. Entirely inside AWS, zero data retention.** All AI runs through AWS Bedrock under a zero-data-retention configuration. Nothing is sent to OpenAI, Anthropic, or any other AI service that could store or train on your data. **3. No logs, no backups, no access.** No logs leave your instance. No backups are taken. SSH is disabled on production instances. There is no remote login, no backdoor, and no support tunnel. Even our team cannot see your data. **4. RAM-only processing, daily teardown.** All files exist only in memory, never written to disk. Every 24 hours, the entire instance is torn down and rebuilt from scratch. Once destroyed, there is nothing for anyone to recover. ### Threat 1: Agentic tools on your desktop are a liability Tools like Claude Cowork and Claude Code are powerful — they are leading the future of AI agents. They're also designed to reach directly into your local filesystem: reading, writing, and modifying files on your machine. That includes whatever happens to be sitting in your Documents, Downloads, or Desktop folders. Contracts. Financial records. Client PII. Tax returns with partial Social Security numbers. Security researchers have already demonstrated that this is not a theoretical risk. In **January 2026, PromptArmor showed** that a hidden prompt injection inside a normal-looking document could silently exfiltrate sensitive files from a Cowork-managed folder, including financial figures and PII, without any user approval. The attack used Anthropic's own API as the outbound channel, bypassing firewalls and sandboxing entirely. Anthropic's own safety guidance warns users to avoid granting Cowork access to financial documents, credentials, or personal records. A security advisory from IRM Consulting is more direct: do not use Cowork for regulated workloads. If your organization handles data covered by SOC 2, HIPAA, or PCI-DSS, these tools should be blocked from any environment that touches regulated data. For law firms, accounting practices, and healthcare organizations, this isn't a matter of configuring permissions carefully. It's a fundamental architecture problem. **The agent shouldn't be on your machine in the first place.** ### Threat 2: The AI providers themselves are a risk OpenAI and Anthropic build the most capable AI models on the planet. That's not in question. What is in question is whether your firm should be sending sensitive client data through their infrastructure. Both companies have experienced significant security incidents in the past year: - **November 2025:** OpenAI disclosed a vendor breach (Mixpanel) that exposed API user data. - **Early 2026:** A ChatGPT data exfiltration vulnerability was patched after researchers demonstrated a hidden channel that bypassed all of ChatGPT's security guardrails. - **March 2026:** Anthropic accidentally leaked the full source code for Claude Code. - **March 2026:** Anthropic's restricted Mythos model was accessed by unauthorized users within hours of being announced. These are the companies building the AI. They are brilliant at it. But every time your team uses ChatGPT or Claude directly, your prompts, documents, and outputs flow through their systems. For firms bound by attorney-client privilege, HIPAA, or fiduciary duty, that's an unacceptable dependency. ### Faradex never touches the AI companies A Faradex instance runs entirely inside Amazon AWS. When your team interacts with AI, those requests go to isolated, zero-data-retention versions of the same frontier models, hosted through AWS Bedrock. Your prompts, responses, and documents are never transmitted to Anthropic, OpenAI, or any other third-party AI provider. They are never stored by Amazon. They are never used to train any model. **You get the same world-class reasoning. None of the exposure.** ### SOC 2 Faradex holds SOC 2 Type I attestation. SOC 2 Type II is targeted for Q3 2026. The security commitments described above are not features waiting on a certification. They are structural properties of the architecture. It is how the system works. --- # AI Workforce — `https://faradex.ai/agentic-ai` ## You're using AI. But you're not using agents. AI is doubling its capabilities every six months. Most professionals think they've kept up because they use ChatGPT or Copilot. They haven't. They've adopted autocomplete. To stay at the forefront, you need AI Agents. ### Where is your firm? Most professionals are stuck in stage two. The frontier has already moved past them. - 30% — No AI - 48% — ChatGPT Era (autocomplete) - 18% — Agents - 4% — Collaborative AI (Faradex synthesis based on data from Microsoft, Salesforce, Deloitte, AMA, PwC, 2024–2026.) ### What is agentic AI? Think of AI agents like employees, not chat windows. With ChatGPT or Copilot, you go back and forth, line by line. You prompt, you read, you refine, you prompt again. You are doing the work. The AI is helping. > Agents flip that relationship. You define the work product you want and send them to go make it. A research memo. A contract comparison. A financial model. You describe the outcome, and the agent delivers it. This is true for individual agent tools like Cowork and Claude Code — though those tools' architecture (reaching into local files) makes them unsafe for regulated work. Faradex takes the agentic model further inside an isolated, secure environment. ### How Faradex uses agents differently Faradex doesn't give you one agent at a time. It gives you a team. Assign a research memo to one, a contract review to another, and a financial summary to a third — all before lunch. They work in parallel, and they keep working even when you close your laptop. Unlike platforms that build proprietary models tuned to a single domain (legal AI, medical AI, finance AI), Faradex takes a different approach. It connects your team to the best frontier models in the world as soon as they ship. Today that's Claude Opus 4.7, the most capable generally available model. Tomorrow it will be whatever comes next. But the models are only part of the equation. Faradex pairs them with a proprietary agent orchestration system that lets your team run dozens of agents at once, in parallel, across different tasks and workstreams. These are frontier models trained on billions of dollars of compute that no single company could replicate on its own, coordinated by an engine built specifically to put them to work for your firm across every subject — not just one narrow specialty. ### ChatGPT / Copilot vs. Faradex | | ChatGPT / Copilot | Faradex | |---|---|---| | Mode | You drive. AI assists. | You direct. Agents execute. | | Concurrency | One conversation at a time. | Many agents run at once. | | Availability | Active only while you're working. | Runs even while you're away. | | Models | General-purpose consumer. | Always the best frontier model. | | Team features | None. | Built for teams. | | Compliance | No compliance story. | Zero data retention. Full control. | | Autonomy | None. | Full. | --- # Team Collaboration — `https://faradex.ai/teamwork` ## AI that's not just a tool, but part of your team. Most AI chatbots and systems are single-user. One person, one chat window, one conversation at a time. Faradex is built for the way professional teams actually work: together, across projects, with shared context. ### Shared project workspaces Every Faradex instance is a shared workspace. Your team members and AI agents all operate in the same context. When someone uploads a document, everyone — human and AI — can reference it. When an agent produces a draft, the whole team can see it, discuss it, and refine it together. No more copying and pasting AI outputs between chat windows and email threads. No more "let me re-explain the context." The AI already knows the project because it's been working on it alongside your team. ### Conversations with AI agents Faradex provides chat rooms where your team can strategize together, with expert tax, finance, and research AI agents in the room. You can work with a client, negotiate with another party, or brainstorm with colleagues — all drawing on AI assistance during the discussion. This isn't a chatbot sitting in a sidebar. These are AI participants in your team's conversations, contributing research, flagging risks, and drafting responses in real time. ### Per-project scope, zero leakage Each project runs on its own isolated AWS instance — separate compute, separate network, separate encryption. There is no shared infrastructure between projects, even within the same firm. Agents have full context within a project but zero visibility into other projects. There is no cross-channel leakage. When the project ends, everything is gone. This means you can run parallel engagements for different clients on the same platform without any risk of information bleeding between them. ### Traditional AI Tools vs. Faradex | | Traditional AI Tools | Faradex | |---|---|---| | User model | Single-user only. | Full team collaboration. | | Context | Resets every conversation. | Persistent, shared project context. | | Workspace | None. | Shared workspace for humans + agents. | | Output sharing | Copy-paste between tools. | Visible to all team members instantly. | | AI role | Sidebar assistant. | Active participant in conversations. | | Project isolation | None. | Per-project AWS instance. Zero leakage. | --- # Article: Your AI Conversations Are Discoverable — `https://faradex.ai/articles/legal-profession-and-ai` ## What Heppner Means for Law Firms Your AI conversations may not be private. Based on a recent district court ruling, they may not be protected by attorney-client privilege. They can be subpoenaed, seized under warrant, and used as evidence — even conversations the provider told you were deleted. If your firm uses ChatGPT, Claude, Copilot, Harvey, or CoCounsel, this article explains what has already happened in court and what it means for your client data. ### Case 1: United States v. Heppner, S.D.N.Y. (Feb. 2026) Bradley Heppner was indicted on federal securities and wire fraud charges. Before his arrest, he used Anthropic's Claude (the free, consumer version) to research legal issues, outline defense strategy, and organize information he had learned from his attorneys. The FBI seized his devices, recovered 31 AI-generated documents, and moved to use them as evidence. Judge Jed S. Rakoff granted the government's motion. The court held that the AI-generated documents were protected by neither attorney-client privilege nor the work product doctrine. **Three findings drove the ruling:** 1. **Claude is not an attorney.** Communications between a client and a non-attorney third party are not privileged. The court found this alone was dispositive. 2. **There was no reasonable expectation of confidentiality.** The court cited Anthropic's privacy policy, which permits the company to collect user inputs and outputs, use them for model training, and disclose them to third parties, including government regulatory authorities. 3. **The work was not directed by counsel.** Heppner acted on his own initiative. Materials generated without attorney direction do not qualify as work product, even if later shared with counsel. This is the first federal district court ruling to squarely address privilege in the context of consumer AI tools. It is a single-judge opinion from the Southern District of New York — persuasive authority rather than binding precedent outside that district, and not yet tested on appeal. That said, the reasoning is straightforward and likely to be followed by other courts considering similar facts. ### Case 2: OpenAI ordered to preserve deleted conversations (May 2025) In *The New York Times Company v. Microsoft Corporation*, a federal court ordered OpenAI to stop deleting ChatGPT conversation logs — including conversations users had already deleted under OpenAI's 30-day retention policy. The court reasoned that deleted chats could contain evidence of copyright infringement. The order affects users of ChatGPT Free, Plus, Pro, and Team, as well as API customers without a Zero Data Retention agreement. OpenAI's COO confirmed the company would comply. **The implication:** "deleted" does not mean gone. If the data existed on the provider's infrastructure at any point, it can be subject to a preservation order, a litigation hold, or a subpoena. ### Case 3: Court orders production of 20 million ChatGPT logs (Jan. 2026) The same litigation produced a far more dramatic order six months later. In January 2026, U.S. District Judge Sidney Stein ordered OpenAI to produce **20 million ChatGPT conversation logs** as evidence in the publishers' copyright claims. The users whose conversations were handed over received no advance notice and had no opportunity to object. The court relied on removing identifying information as a privacy safeguard. But anonymization of AI conversation logs has already proven unreliable. When *The Washington Post* examined 47,000 leaked ChatGPT logs, analysts found that email addresses, phone numbers, and intimate personal details remained visible in the data, making re-identification straightforward despite name removal. This is the trajectory: first the court said "stop deleting." Then the court said "hand over 20 million conversations to the opposing party." The direction is clear, and it is accelerating. Any data that exists on a provider's infrastructure is reachable, and the courts are increasingly willing to reach for it. ### The CLOUD Act makes this worse The Clarifying Lawful Overseas Use of Data Act (2018) requires any provider of electronic communication services or remote computing services to comply with a U.S. law enforcement warrant or subpoena to produce data within its "possession, custody, or control" — regardless of where that data is physically stored. The CLOUD Act applies when three conditions are met: 1. A U.S. court has jurisdiction over the entity. 2. The entity qualifies as an electronic communication or remote computing service provider. 3. The entity has possession, custody, or control over the data being sought. Every major AI platform — OpenAI, Anthropic, Google, Microsoft, Harvey, CoCounsel — meets all three conditions. They are U.S.-based or U.S.-subject service providers. They retain user data for some period. They have possession and control of that data during the retention window. Harvey AI has publicly discussed this exposure, noting that its short retention windows reduce the practical risk of a CLOUD Act demand. That may be true as a practical matter, but **reduced risk is not eliminated risk**. During any retention window (Harvey's minimum is three hours), the data exists, the provider controls it, and it is reachable by warrant. ### "But we use enterprise tools, not free ChatGPT" The most common reaction to Heppner from law firms is: "That case involved a criminal defendant using free consumer AI without his lawyer's involvement. We use paid enterprise platforms with confidentiality agreements. It doesn't apply to us." This distinction is partially legitimate on its facts. But it does not provide the comfort firms think it does. **What is different about enterprise use:** - Heppner was a client acting alone, without attorney direction. When a lawyer uses an AI tool as part of active case preparation, the work product analysis is stronger. - Enterprise agreements typically contain explicit confidentiality obligations that consumer terms of service do not. This helps on the "reasonable expectation of confidentiality" prong. - Enterprise platforms like Harvey operate under shorter retention windows and stricter data handling commitments than consumer tools. **What is not different:** - The court's first holding — that an AI platform is a non-attorney third party — has nothing to do with price or subscription tier. Harvey is not an attorney. CoCounsel is not an attorney. Claude Enterprise is not an attorney. **Paying more does not transform software into a member of the bar.** - Enterprise platforms still retain data, even briefly. They still have technical access to it during processing. Their privacy policies still reserve the right to disclose under legal process. A contractual promise of confidentiality improves your position, but courts look at the totality of circumstances. - The work product doctrine protects against discovery by an adversary. It does not protect against subpoenas directed at the platform itself. Even if your work qualifies as attorney work product, a court can still compel the platform to produce it in response to a third-party subpoena, as the OpenAI production order demonstrates. **The honest assessment:** Heppner involved the worst possible facts for privilege. Your firm's facts are better. But the structural vulnerability remains. Enterprise agreements reduce your exposure on the confidentiality prong but do not eliminate it. The only way to fully close the gap is to ensure no data exists to be compelled in the first place. ### Why Faradex is different Every platform discussed above shares a common vulnerability: **they write your data to disk.** Whether they keep it for three hours or thirty days, that act of persistence is what creates legal exposure. Once data hits a filesystem or database on the provider's infrastructure, it can be subpoenaed, preserved under court order, or compelled under the CLOUD Act. The retention window is a question of degree, not kind. > Faradex is fundamentally different. It is, as far as we are aware, the only AI platform built from the ground up so that your data never touches persistent storage. Everything runs in volatile memory. Nothing is written to disk. Faradex maintains a database, files, and records of your conversations while you are working — but all of it lives exclusively in RAM. There is no disk write, no filesystem persistence, no database commit to durable storage. When your session ends, that memory is released. The data does not get deleted. It was never stored in the first place. **This is not a retention policy. It is a hardware-level architectural constraint.** **Why this matters legally:** A subpoena or CLOUD Act demand can only compel production of data within the provider's "possession, custody, or control." Data that exists only in volatile memory during active processing, and is never written to any persistent medium, is not data that Faradex or Amazon Web Services could retrieve and hand over after the fact. There is no backup to restore, no database to query, no log to pull. The architecture makes compliance with a production demand for historical conversation content a factual impossibility, not a policy choice. This is not how other platforms work. Most AI providers write your inputs and outputs to disk the moment they arrive, then promise to delete them later. Some promise deletion after 30 days. Some after three hours. But the act of writing to persistent storage is itself the problem: it creates data that can be frozen by a litigation hold, reached by a warrant, or preserved under court order before the deletion window closes. **Faradex never creates that data in the first place.** **AI providers never receive your data in a retainable form.** Faradex operates under Zero Data Retention agreements with upstream AI providers. Under these agreements, providers do not store your inputs or outputs and do not train on your data. When the API call completes, nothing remains on their infrastructure under the terms of those agreements. **Dedicated, single-tenant infrastructure.** Your Faradex environment runs on its own isolated AWS instance. No shared infrastructure, no multi-tenant commingling, no lateral exposure. Each project within your account also runs on its own isolated AWS instance — a separate server, not a logical partition within a shared cluster. **We can be subpoenaed. We designed the system so there is nothing responsive to produce.** Faradex is a U.S.-based service provider running on AWS. We are subject to the same legal process as any other company. A court could issue a warrant, a subpoena, or a CLOUD Act order directed at Faradex, and we would be obligated to comply. The difference is what compliance looks like. For the substance of your legal work — your prompts, your documents, the AI's analysis — there is nothing to produce. That content existed in RAM during your session and nowhere else. It cannot be recovered after the fact by us or by Amazon. > Other platforms argue they are unlikely targets. We don't make that argument. We assume we will be targeted. We built the architecture so it does not matter. For lawyers specifically, this architecture solves the Heppner problem at its root. The court in Heppner found no reasonable expectation of confidentiality because Anthropic's privacy policy allowed collection, retention, and disclosure of user data. Faradex cannot collect, retain, or disclose conversation content, because the system is physically incapable of doing so. **You cannot subpoena what does not exist. You cannot preserve what was never stored.** ### What this means for your firm The Heppner ruling is a strong signal (though not yet appellate precedent) that AI platforms are third-party intermediaries whose privacy policies do not create a reasonable expectation of confidentiality sufficient to support privilege. The OpenAI orders confirmed the corollary: data you believe is deleted may still exist, and courts are willing to compel its production on a massive scale without notifying the affected users. For law firms, the exposure is not limited to criminal defense. Any client matter processed through a consumer or enterprise AI tool — contract review, M&A due diligence, IP analysis, regulatory research — creates data that could be discoverable in litigation, subject to a preservation hold, or producible under subpoena. The "we use enterprise tools" defense narrows the exposure but does not close it. Faradex was built for firms that cannot accept that residual risk. Your prompts, your documents, your outputs — the substantive content of your legal work exists in volatile memory during processing and nowhere else. When the work is done, the data is gone. Not because we promise to delete it. Because it was never written to disk. ### Sources cited on this article Proskauer · Gibson Dunn · Harvard Law Review · Malwarebytes · SiliconANGLE · GBlock · Congress.gov (CLOUD Act text) · AWS CLOUD Act FAQ · Harvey blog on data sovereignty. --- # Article: Private AI Is a Human Right — `https://faradex.ai/articles/private-ai-human-right` We are approaching something close to omniscience. A system that knows everything humans have ever written, can reason across all of it, and can apply that reasoning to any question you ask. Medicine, law, finance, philosophy. On demand, in seconds. > **Right now, every time you access AI, someone is recording what you said.** Every major AI platform stores your conversations. Their privacy policies grant broad rights to that data: collection, training, disclosure to governments. In February 2026, a federal court ruled that a user's AI conversations could be seized by the FBI and used as evidence against him. The court found there was no expectation of privacy. A few months before that, a court ordered OpenAI to preserve ChatGPT conversations that users had already deleted. **This is not a future risk. It is current case law.** Now think about where AI is headed. The system you talk to next year will know more than any human who has ever lived. It will reason about your health, your legal exposure, your finances, your ideas. People will use it the way they use their own minds. To think through hard problems. To explore uncomfortable questions. To process things they wouldn't say out loud. If every one of those thoughts is recorded, stored on someone else's server, and available to anyone with a subpoena, people will self-censor. The most powerful tool in human history will be crippled by the reasonable fear that someone is watching. > **Accessing AI can't come at the expense of privacy.** For all of human history, your thoughts were private. Not because of any law. Because of physics. Nobody could read your mind. AI changes that, because it is becoming the place where people think. When you externalize your reasoning to an AI, and that AI keeps a record, your inner life is no longer yours alone. > **We believe that access to AI, without a record of what you asked, is a fundamental human right.** Not a premium feature. Not an enterprise add-on. Not a toggle in your settings. A right. This is a founding principle of Faraday Collective. Everything we build starts here. --- # In the News — `https://faradex.ai/news` Curated reporting and coverage that motivates Faradex's architecture — federal discovery rulings on AI conversations, court-ordered preservation of deleted logs, breaches and credential leaks at major AI providers, regulator guidance on AI under HIPAA and attorney-client privilege, and enterprise incidents involving consumer AI tools. ## Discovery, preservation, and the new evidentiary record - "OpenAI ordered to share users' personal info with Meta and Google" — Futurism. https://futurism.com/artificial-intelligence/openai-personal-information-meta-google - "OpenAI ordered to produce 20M ChatGPT conversations to the NYT" — Ars Technica. https://arstechnica.com/tech-policy/2025/11/openai-fights-order-to-hand-over-20-million-private-chatgpt-conversations/ - "Federal court forces OpenAI to retain all ChatGPT logs — including deleted ones" — Adweek. https://www.adweek.com/media/openai-fights-court-order-requiring-it-to-store-deleted-chatgpt-conversations-indefinitely/ - "Everything you typed into ChatGPT may now be legally preserved" — PMG360. https://pmg360.com/blog/everything-you-typed-into-chatgpt-may-now-be-legally-preserved - "AI chats are now discoverable evidence (NYT v. OpenAI)" — Terms.law. https://www.terms.law/2025/11/12/openai-v-new-york-times-stopped-being-just-a-copyright-case-the-moment-the-court-turned-to-your-chatgpt-logs/ ## Breaches, leaks, and vulnerabilities at AI providers - "Anthropic source code leaks via Claude Code — second security lapse in days" — Fortune. https://fortune.com/2026/03/31/anthropic-source-code-claude-code-data-leak-second-security-lapse-days-after-accidentally-revealing-mythos/ - "Anthropic confirms it suffered a data leak" — VentureBeat. https://venturebeat.com/ai/anthropic-confirms-it-suffered-a-data-leak - "OpenAI was hacked in 2023 — and didn't disclose it" — New York Times. https://www.nytimes.com/2024/07/04/technology/openai-hack.html - "ChatGPT bug leaked conversation histories and payment data" — BBC. https://www.bbc.com/news/technology-65047304 - "100,000+ stolen ChatGPT credentials sold on the dark web" — The Hacker News. https://thehackernews.com/2023/06/over-100000-stolen-chatgpt-account.html - "Claude code-interpreter vulnerability exposes enterprise data" — CSO Online. https://www.csoonline.com/article/4082514/claude-ai-vulnerability-exposes-enterprise-data-through-code-interpreter-exploit.html ## Enterprise incidents and behavioral data - "Samsung bans ChatGPT after engineers leak source code three times in 20 days" — TechCrunch. https://techcrunch.com/2023/05/02/samsung-bans-use-of-generative-ai-tools-like-chatgpt-after-april-internal-data-leak/ - "11% of data employees paste into ChatGPT is confidential" — Cyberhaven. https://www.cyberhaven.com/blog/4-2-of-workers-have-pasted-company-data-into-chatgpt - "Italy temporarily bans ChatGPT over privacy concerns" — New York Times. https://www.nytimes.com/2023/03/31/technology/chatgpt-italy-ban.html ## Privilege and regulated-industry guidance - "New ChatGPT privacy concerns for lawyers and legal staff" — Bar Association of San Francisco. https://www.sfbar.org/blog/heads-up-new-chatgpt-privacy-concerns-for-lawyers-and-legal-staff/ - "Careless generative AI use puts attorney-client privilege at risk" — Bloomberg Law. https://news.bloomberglaw.com/bloomberg-law-analysis/careless-generative-ai-use-puts-attorney-client-privilege-at-risk - "ABA: privilege risk when transmitting client data to AI providers" — American Bar Association. https://www.americanbar.org/groups/litigation/resources/newsletters/corporate-counsel/ai-attorney-client-privilege/ - "Navigating AI compliance with HIPAA — consumer tools fail the bar" — Norton Rose Fulbright. https://www.nortonrosefulbright.com/en-us/knowledge/publications/55f5440a/navigating-ai-compliance-with-hipaa-essentials ## Copilot, Microsoft 365, and shared-tenant exposure - "Gartner: five security risks in Microsoft 365 Copilot" — WinBuzzer. https://winbuzzer.com/2026/03/17/gartner-five-security-risks - "Microsoft Copilot exposed Fortune 500 data via permission inheritance" — Lasso Security. https://lasso.security/blog/microsoft-copilot - "'BioShocking' attack tricks AI browsers into stealing credentials" — SecurityWeek. https://www.securityweek.com/bioshocking-attack-tricks-ai-browsers-into-stealing-credentials/ --- # Company - **Legal entity:** Faraday Collective Inc. - **Product:** Faradex - **Infrastructure:** AWS only. All inference via AWS Bedrock under zero-data-retention agreements. No data leaves the USA. - **Isolation model:** dedicated, single-tenant AWS instance per customer. Each project within a customer account also runs on its own isolated AWS instance — separate compute, network, and encryption. - **Current frontier model:** Claude Opus 4.7. Frontier-model access updates as new models ship. No proprietary lock-in. - **Compliance:** SOC 2 Type I attested. SOC 2 Type II targeted Q3 2026. Security properties described are structural (architectural), not feature flags pending certification. - **Target customers:** law firms, accounting firms, investment funds, and other professional services teams under attorney-client privilege, HIPAA, SOC 2, PCI-DSS, or fiduciary obligations. - **Posture on legal demands:** Faradex is a U.S.-subject company and can be subpoenaed. The architecture is designed so compliance with a production demand for substantive content is a factual impossibility — that content exists only in RAM during active processing. *You cannot subpoena what does not exist. You cannot preserve what was never stored.*