Built in West Oakland,
Forged in High-Stakes Environments
Ciph Lab began when I watched enterprise after enterprise struggle with the same hidden problem: AI was accelerating faster than the organizational structures built to manage it. When AI does the work, someone still has to answer for it, and in most organizations no one does.
13+ years running legal and IP operations at Amazon Lab126, Apple, Okta, and Figma, after six years at a patent boutique. 5,500+ patent applications managed, including ~1,400 a year and an estimated $100M+ in outside counsel spend at Amazon, in work where one missed handoff can't be undone.
AI agents now act on behalf of people across legal and enterprise work, but the ownership structures underneath them were built for humans. Physical AI companies are now leasing space in West Oakland, where Ciph Lab is headquartered.
IR™ Continuous Ownership, a working prototype starting with IP work; the GRACI™ protocol (patent pending); a completed beta cohort of the IR™ AI Governance Operator Certificate; and a Delaware public benefit corporation.
Headquartered in West Oakland,
Where the West Coast Standardized the Handoff
Ciph Lab is headquartered in West Oakland, at the center of the Bay Area's AI map: one BART stop from downtown San Francisco, where much of enterprise AI is decided, and a short trip from UC Berkeley, whose research labs and graduates feed the next generation of AI talent.
This neighborhood has standardized a handoff before. In 1962, the Port of Oakland became the first major port on the West Coast to build container terminals and invest in container cranes. Containerization replaced slow, manual cargo handling with a standard unit that moved through every handoff the same way, every time, and it changed global trade.
That change came at a real cost to West Oakland. Urban renewal, freeway construction, and BART displaced thousands of families in what was then a predominantly Black neighborhood, and port truck traffic still runs through its streets. Progress that moves through a community without serving it is not a pattern Ciph Lab intends to repeat.
Now West Oakland is becoming a home for physical AI. In August 2026, Jeff Bezos's physical AI startup Prometheus and Elon Musk's Tesla both signed leases at the American Steel Blocks on Mandela Parkway, as reported by the San Francisco Business Times and KALW. As AI moves off the screen and into factories, labs, and supply chains, the question of who answers for what it does gets harder, not easier.
Legal and AI-assisted work still have no standard handoff. Documents move between internal teams, AI tools, and outside counsel, and named responsibility falls away at every step. Ciph Lab is building that standard.
The same gap runs through this neighborhood. When the port standardized the handoff in 1962, West Oakland was never treated as a party to it. The cargo moved efficiently, and the community absorbed the cost. Now physical AI is moving onto Mandela Parkway, and the same question returns: when an AI system acts on a street, in a warehouse, or in a city decision, who answers for it, and who are they answerable to?
Ciph Lab's answer is the same for Oakland as it is for the enterprise: every handoff has a named owner, and the people affected can see who that is. And because accountability is work, not just software, we intend for the people who hold those roles to come from here. Containerization automated jobs out of West Oakland. AI accountability can bring them back.
As a public benefit corporation, this is not a side project. It is the standard we're building, applied first to the place we're building it.
Companies Deploy AI —
Then Realize They Weren't Ready
Across multiple high-scale tech companies — from product development to enterprise SaaS — I watched the same pattern unfold: organizations would roll out AI tools, and then everything would become messy. Teams would realize too late that the underlying organizational structures weren't designed for AI. Employees lacked the skills to use it effectively. No one clearly owned how AI was used, or answered for what it produced.
The chaos wasn't the AI itself — it was deploying intelligent systems into organizations that weren't ready for them. And retrofitting readiness after deployment is exponentially harder than building it in from the start.
Everyone was racing to adopt AI without asking the foundational question: Is our organization actually prepared to support this?
Learning Precision in
High-Stakes Environments
My career began at Haley Guiliano, a Ropes & Gray spinoff, where I spent six years as a Senior Patent Paralegal managing global operations for tech giants including Google. Managing 500+ complex jurisdictional filings taught me something fundamental:
"In high-velocity environments, a 0.1% error rate in the foundation leads to 100% failure at scale."
One missed deadline in patent law doesn't just delay a project — it can invalidate years of R&D investment. One misrouted document doesn't slow things down — it creates legal exposure that compounds. This "zero-defect" culture became the lens through which I viewed every system I touched afterward.
Later, over three and a half years at Amazon Lab126, I managed ~1,400 patent applications a year (5,000+ in total) and 10+ outside law firms, overseeing an estimated $10M in outside counsel spend per firm (more than $100M in total), while building operational frameworks that bridged product innovation with legal and compliance requirements. That's when I realized: the companies that scale successfully don't retrofit compliance — they architect it into the foundation.
The question isn't whether AI will act on your organization's behalf. It's whether anyone will still answer for it when it does.
Every Company Has a Security Model.
None of Them Were Built for AI.
Who answers for an AI system's actions depends on something most AI conversations skip: how the company already decides who can access what, and who has to approve it. That is the structure AI inherits the day it's deployed.
Over 13+ years, from a patent boutique serving clients like Google to Apple, Amazon Lab126, Okta, and Figma, I built legal operations infrastructure inside organizations with fundamentally different approaches to risk, access, and security. There isn't one way companies operate. There are at least six distinct models. Each assigns responsibility differently, and each breaks differently once AI starts acting inside it. Not one of them was designed with AI agents in mind.
That is why Ciph Lab doesn't sell one accountability playbook. An approach that only works in one model fails in the other five. We start by identifying which model you operate in, because that determines where ownership will fall away first.
Access is presumed until proven problematic. Incredible velocity — teams move fast, experiment freely, innovate without friction. But requires strong oversight to catch problems before they compound. AI can move fast here — but without clear ownership, it moves dangerously.
Permissions granted layer-by-layer. Risk reduced dramatically. But when the C-suite mandates AI transformation as a company goal, this model creates real execution friction. The intent is there. The infrastructure wasn't built for it. Not impossible — but it requires deliberate redesign before deployment, not after.
Every request verified continuously — no implicit trust, even inside the network. More rigorous than restrict-first. AI deployment here requires proving every model call, every data touch, every agent action meets security standards in real time. Few organizations are equipped for this.
No central authority owns access decisions — individual business units decide on their own AI tools within loose corporate guidelines. Shadow AI explodes because there's no single front door. Every department becomes its own island, making isolated decisions that create enterprise-wide risk.
Decisions driven entirely by external regulation — HIPAA, SOC 2, FedRAMP, GDPR. AI deployment gets evaluated through "does this pass audit?" rather than "are we organizationally ready?" Compliance and readiness are not the same thing — and companies confuse them constantly.
Default is open, restrictions added reactively after something goes wrong. Velocity is maximum. Accountability stays minimal until a breach, a PR crisis, or a regulatory action forces a reckoning. Many growing companies start here and realize too late they need infrastructure retroactively.
Companies don't fail at AI because they picked the wrong security model. They fail because they deploy before assessing whether their organization — whatever its structure — is actually ready to support AI. The model you operate in shapes how you get ready. It doesn't determine whether you have to.
None of These Models
Were Built for the Agentic Era.
Here's the point most conversations about AI miss entirely: every security model described above — trust-and-verify, restrict-first, zero trust, federated, compliance-led, permissive — was designed for human access to data systems. A person requesting access. A person making a decision. A person responsible for an outcome.
AI agents change that equation completely. Agents don't request access — they act. They don't review information — they make decisions and execute on them. They don't wait for approval — they operate continuously, often invisibly, across systems that were never designed to track non-human actors.
This means the accountability gap isn't just about deploying AI tools carefully. It's about the fact that the organizational infrastructure underneath AI — the access models, the approval workflows, the ownership structures — was never built to hold anyone answerable for systems that act on behalf of humans. And it breaks in different ways depending on which model you operate in.
Intelligence Resources™ is not a compliance layer you place on top of an existing security model. It is a rethink of the organizational structure underneath it. A new corporate function — like HR for people or IT for systems — that permanently owns the question of how AI operates inside your organization. Not as a one-time project. As a lasting capability.
Agents inherit presumed access. No mechanism exists to catch what they do with it. Accountability arrives after the damage.
Approval workflows were designed for human requests. Agents generate thousands of micro-decisions no approval process can track in real time.
Agents cross departmental boundaries invisibly. Audits catch what already happened. Neither model was built for continuous, cross-functional AI accountability.
Everyone Was Asking
the Wrong Question
Everywhere I worked, leadership asked some version of the same question: "How do we deploy AI faster?"
The better question is: "Are we ready for what happens after we deploy it?"
That gap — between deployment speed and organizational readiness — is where Innovation Teams burn resources cleaning up post-deployment chaos, where Compliance becomes an emergency response team instead of a strategic partner, and where Legal Operations scrambles to create guardrails that should have existed from day one.
Organizations don't need faster AI deployment. They need to assess whether they have the Intelligence Resources™ to support AI sustainably — before things get messy.
Intelligence Resources™:
Readiness Before Deployment
I founded Ciph Lab in West Oakland to solve the root cause: enterprises lack a systematic way to assess AI readiness before deployment — so they roll out tools into unprepared organizations and spend years managing the aftermath.
Intelligence Resources™ is a structured methodology that diagnoses organizational readiness before AI tools go live, then teaches organizations how to redesign their operations for the AI era. This isn't about accepting your constraints — it's about understanding your starting point so we can design the right transformation for your context.
That transformation includes fundamentals most companies overlook: what new roles you need to create, what skills to hire for, how decision rights should change, how workflows must evolve, and what accountability structures actually work in practice. Intelligence Resources™ isn't a framework you bolt onto existing operations — it's a new corporate function that teaches organizations how to operate differently in the AI era.
My approach combines a B.S. in Legal Studies, a UC Berkeley certificate in business administration, and an MBA completing in December 2026 with hands-on experience building operational frameworks across different security philosophies in high-stakes, highly-regulated environments where precision isn't optional.
From Readiness
to Continuous Ownership
A readiness assessment tells an organization where it stands on day one. It doesn't keep anyone answerable on day ninety. That's why Ciph Lab's core build is IR™ Continuous Ownership: a working prototype that keeps a named owner attached to AI-assisted work at every handoff.
We're starting with IP and legal work, where I spent my career, where a dropped handoff is most expensive, and where the cost is easiest to measure. From there, the same standard extends to any work AI touches, including the physical AI now arriving in our own neighborhood.
"Ciph Lab exists because the gap between AI deployment and organizational readiness is too expensive to ignore — and too predictable to keep repeating."
Ready to find out where you stand?
Get your free AI Intelligence Score™ — a 10-minute diagnostic that shows exactly where your organization is ready and where it isn't, regardless of which model you operate in.
Investing in AI accountability? See the investor overview or email founder@ciph-lab.com.