MorphicBrain AI seeks one enterprise partner to test a different protection question: what happens to consequential work when AI is already acting?
LAGUNA NIGUEL, Calif., Sept. 30, 2026 /PRNewswire-PRWeb/ -- As NVIDIA and other major technology companies build increasingly sophisticated ways to govern, constrain and secure autonomous AI agents, Daniel Nicolas is asking a different question:
WHAT HAPPENS WHEN THE AI STAYS WITHIN ITS AUTHORITY, BUT THE CONSEQUENTIAL WORK ITSELF BEGINS TO GO WRONG?
Nicolas, 83, is the founder of MorphicBrain AI, LLC and the developer of Josephine, a patent-pending approach designed around a simple proposition: stopping, constraining or quarantining an AI agent is not necessarily the same thing as protecting the work the agent has already put into motion.
"I didn't set out to compete with NVIDIA, or with anyone else," Nicolas said. "I started with a problem that bothered me. AI is beginning to do real work — payments, claims, instructions, approvals and other consequential activities. If something starts going wrong, stopping the AI may be important. But what happened to the work?"
PROTECT THE WORK, NOT JUST THE AI
Josephine is being developed as a protective layer that travels beside consequential AI-driven work.
Rather than attempting to replace existing AI security, governance or human decision-making, Josephine focuses on evidence describing the work itself.
Within boundaries established in advance by the organization, its purpose is to recognize specified deterioration conditions and invoke only pre-authorized protections intended to preserve management reaction time before consequences become difficult or impossible to reverse.
The company does not claim that Josephine has yet been proven in a live enterprise deployment.
Josephine has undergone controlled local and synthetic testing, and MorphicBrain AI is now seeking one organization willing to conduct a bounded, supervised pilot against real organizational evidence.
The objective is not to demonstrate that Josephine works by definition. It is to find out whether the proposition survives contact with real consequential work.
A DIFFERENT LAYER OF THE PROBLEM
Recent developments in agent security make the distinction increasingly relevant.
Agent-level security can answer questions such as: What may this agent access? What actions may it take? Can those permissions be independently enforced? Can the agent be isolated or stopped if it exceeds them?
Josephine asks another question:
WHAT IS HAPPENING TO THE WORK ITSELF?
A payment, instruction, approval, claim or other business process may already be moving through downstream systems.
An agent can remain within its assigned permissions while the work nevertheless moves toward an unwanted result.
Josephine is intended to operate at that work layer — not as a replacement for agent containment, but potentially beside it.
EVIDENCE THAT PROTECTION ACTUALLY REACHED THE WORK
Work on Josephine has also exposed another problem: evidence that protection was requested is not necessarily evidence that protection succeeded.
That led Nicolas to preserve a sequence that keeps five states separate:
REQUESTED → ACKNOWLEDGED → APPLIED → CONSUMED → EFFECT OBSERVED
The distinction matters because a system should not claim that consequential work was protected merely because somebody requested an intervention or another system acknowledged it.
An experimental future extension, the Josephine Protection Receipt (JPR), has now completed its bounded laboratory capability phase.
JPR is designed to preserve an evidence-grounded account of what happened to the work and to the protection afterward — including what remains unresolved or cannot be established from available evidence.
JPR remains separate from the frozen, pilot-ready Josephine MVP. It has not been integrated into that MVP and is not part of the current pilot proposition.
AUDIT THE AUDITOR
The same evidence discipline produced another principle Nicolas calls "Audit the Auditor."
An external certification, validator, monitoring system, vendor assertion or auditor conclusion may itself become part of the evidence surrounding consequential AI work.
Josephine's experimental work does not assume that the existence of such an assertion automatically establishes the underlying fact.
The question remains:
WHAT DOES THE EVIDENCE ACTUALLY ESTABLISH?
Nicolas deliberately stopped expanding JPR when testing indicated that going further would risk turning it into something it was not intended to be — a truth engine, causal engine, decision-maker or another protection system.
BUILT WITHOUT A CONVENTIONAL AI COMPANY BEHIND IT
Nicolas is not a software engineer and did not build Josephine with a conventional development team or venture-backed budget.
He developed the system iteratively with the assistance of AI, while maintaining a strict distinction between what testing has demonstrated and what remains to be proven.
At 83, he sees that fact not as the technical argument for Josephine, but as part of a broader change in who can investigate difficult problems with AI assistance.
"The interesting thing isn't that I am 83," Nicolas said. "The interesting thing is that someone like me can now pursue a problem far enough to build something testable. But AI helping me build it does not prove that it works. That's why I need the pilot."
SEEKING ONE CONTROLLED PILOT
MorphicBrain AI is seeking one organization with a consequential AI-assisted or agentic workflow willing to define a bounded supervised pilot.
Before the pilot begins, the parties would agree on the work being protected, the evidence available, the organization's authority boundaries, the conditions that would constitute deterioration, the protections the organization authorizes, and what would count as success or failure.
Josephine would not receive open-ended authority.
Management would remain in control.
And failure would be evidence too.
"If Josephine cannot create useful reaction time or cannot establish from evidence what happened to the work, I want to know that," Nicolas said. "The next step isn't another feature. The next step is evidence from a real organization."
Josephine has been developed by MorphicBrain AI, LLC in Laguna Niguel, California.
The underlying work is the subject of a U.S. provisional patent application covering systems and methods for traveling, evidence-driven protection, verification and recovery of consequential AI-driven work.
More information about Josephine and the controlled-pilot proposition is available at josephine.asiveritas.com.
Video: https://josephine.asiveritas.com/media/Josephine_Pilot_Overview.mp4
Media Contact
Daniel Nicolas, Morphicbraqin AI, LLC, 1 949 933 0800, [email protected], josephine.asiveritas.com
SOURCE MorphicBrain AI
Share this article