Dr. Freddie Seba

AI Governance Keynote Speaker  ·  Author  ·  Scholar Operator

EdD · USF  ·  MBA · Yale  ·  MA · Stanford  · Teaching · UIC  ·  20+ years · Silicon Valley Founder & Global Executive · Digital Health · Fintech · Higher Ed.

Issue #67 | AI Governance Is Entering Its Control Phase

Compute, Talent, Agents, Insurance, Education, Healthcare, and the New Institutional Test

AI Ethics & Governance for Leaders, Boards & Trustees

By Dr. Freddie Seba

© 2026 Freddie Seba. All rights reserved.

Editorial Note

Over the past several weeks — across recent AI convenings in Silicon Valley, my ongoing AI governance book work, and the themes developed in Issues #65 and #66 — I have been tracing a shift in AI governance from adoption and experimentation toward autonomy, discernment, and institutional control.

Over the past two issues, I have been tracing a shift in AI governance.

Issue #65 focused on autonomy: AI systems are moving from assistive tools into live workflows.

Issue #66 focused on discernment: capability is accelerating faster than many institutions can judge, challenge, or contain it.

This week’s research sharpens the next lesson: governance must now become an operating discipline of control.

The issue is no longer only whether AI is powerful. The question is whether institutions can control the conditions under which that power enters into work.

This week’s signals came from many directions:

  • Google’s reported multibillion-dollar investment in Anthropic
  • DeepSeek’s new frontier-challenging model preview
  • Talent movement from Meta into the Thinking Machines Lab
  • OpenAI’s expanded Codex capabilities
  • Anthropic’s staged cyber-safeguard strategy
  • Stanford HAI’s work on LLMs for workplace social skills
  • MIT’s call for more “humble” AI
  • Education-sector debates over bans versus governed enablement
  • AI-generated music flooding platforms
  • Health AI evidence and uncertainty frameworks
  • Insurance carriers are beginning to exclude AI-related harms

Different sectors. Same governance signal.

AI is becoming infrastructure, labor, an interface, a content engine, a clinical assistant, a coding partner, a cyber variable, and a strategic asset at the same time.

That means governance has to move from policy language to operating control.

This Week’s Governance Lesson

Control is governance after capability.

  • Capability creates pressure.
  • Control creates legitimacy.

Boards and executive teams should not mistake model progress, vendor partnerships, or employee adoption for institutional readiness.

The real question is whether the organization can explain:

  • Where AI is being used
  • What AI is allowed to do
  • What AI can access
  • Who owns the use case
  • What evidence supports deployment
  • What happens when the system fails
  • Who can pause, restrict, or stop use

That is the practical difference between AI adoption and AI governance.

Executive Reflection

AI Is Moving Faster Than Ownership

This week’s strongest leadership signal came from the ownership gap.

Fortune reported on a Pearl Meyer survey showing that board members and executives do not fully agree on who owns AI strategy and implementation. Board members emphasized C-suite ownership, while executives distributed responsibility across the C-suite, business leaders, and functional heads. That gap matters.

Distributed involvement is not the same as distributed accountability.

AI now touches nearly every part of the enterprise:

  • Product
  • Operations
  • Legal
  • HR
  • Finance
  • Cybersecurity
  • Procurement
  • Customer service
  • Education
  • Clinical practice
  • Research
  • Workforce redesign

So it is understandable that many executives see AI as cross-functional.

But from a board perspective:

“Everyone touches it” can quickly become “no one owns it.”

The governance failure comes when the board hears the AI strategy story but does not see the operating model underneath it.

That is the issue this week.

Not AI use. AI control.

What We Are Seeing: Signals

1. Compute Is Becoming Strategy — and Governance Exposure

Google’s reported plan to invest up to $40 billion in Anthropic is not just a financing story. It is a signal that frontier AI is increasingly defined by access to infrastructure:

  • Chips
  • Cloud capacity
  • Energy
  • Data centers
  • Deployment scale
  • Strategic compute partnerships

AI governance translation:

Computing is no longer a back-office technology issue. It is a strategic dependency.

Board Move:

Ask management which AI capabilities depend on which vendors, clouds, chips, data centers, contracts, and exit options.

Vendor concentration is now governance exposure.

2. Frontier Competition Is Narrowing the Comfort Zone for Institutional Delay

DeepSeek’s preview of new models points to another governance reality: frontier capability is becoming more competitive globally.

Whether every benchmark claim holds under independent review is not the only point. The governance signal is that capability gaps are narrowing, model options are multiplying, and institutions will face pressure to adopt powerful systems from multiple vendors, jurisdictions, and infrastructure ecosystems.

AI governance translation:

Model competition increases choice, but it also increases the due diligence burden.

Board Move:

Require an AI vendor and model provenance review that covers:

  • Jurisdiction
  • Data handling
  • Benchmark evidence
  • Security posture
  • Update cadence
  • Contractual protections
  • Exit readiness

3. Talent Movement Shows That AI Advantage Is Organizational, Not Just Technical

TechCrunch’s coverage of Thinking Machines Lab shows how quickly AI talent, infrastructure, and capital are being reorganized. This matters because institutional leaders often treat AI strategy as a tool-selection problem.

  • It is not.
  • It is a capability-system problem.

AI advantage depends on:

  • Human talent
  • Compute
  • Data access
  • Workflow integration
  • Vendor relationships
  • Safety posture
  • Governance maturity

AI governance translation:

AI advantage is not just model access. It is organizational capacity.

Board Move:

Ask whether the institution has the internal capability to evaluate, integrate, monitor, and challenge AI systems — or whether it is simply renting a capability it cannot govern.

4. Agents Are Moving From Assistance to Computer Use

OpenAI’s Codex update is a clear signal that AI agents are moving deeper into software work and operating environments.

When AI can operate across applications, remember preferences, and carry work forward, oversight has to shift.

  • The question is no longer only: What prompt did the user enter?
  • The better question is: What operating authority has the system been granted?

AI governance translation:

Agentic capability turns acceptable-use policy into access-control governance.

Board Move:

Require agent permissions to be:

  • Tiered
  • Logged
  • Reviewable
  • Reversible
  • Tied to named human owners

5. Cyber Capability Is Becoming a Staged-Release Governance Problem

Anthropic’s Opus 4.7 announcement matters not only because of performance claims, but because of its release posture.

Anthropic described testing cyber safeguards on less capable models before the broader release of more advanced Mythos-class models.

At the same time, public debate around Mythos shows how difficult cyber-risk communication has become.

Both points can be true:

  • Cyber capability may be genuinely serious.
  • Frontier labs may also use risk narratives to shape markets, access, and trust.

That is why boards need governance translation, not hype translation.

AI governance translation:

Cyber-risk claims require independent review, not automatic acceptance or dismissal.

Board Move:

Ask management:

  • How are frontier-model cyber claims being evaluated?
  • Who validates vendor safety representations?
  • What internal cyber controls have changed because of AI capability gains?

6. AI Is Entering the Social Fabric of Work

Stanford HAI highlighted research using LLMs to help people practice workplace-relevant social skills, including conflict resolution, peer counseling, and novice therapy skills.

This is promising. It is also governance-intensive.

When AI becomes a practice partner for empathy, conflict resolution, counseling, teaching, caregiving, or management, the institution is not only deploying a productivity tool. It is shaping interpersonal norms.

AI governance translation:

Workplace AI can train behavior, not just automate tasks.

Board Move:

Require review of AI tools that simulate human interaction, coach employees, or influence social-emotional behavior — especially in education, healthcare, financial services, HR, and youth-facing settings.

7. Healthcare AI Needs Humility, Not Authority Theater

MIT News summarized work warning that AI systems can steer clinicians in the wrong direction when they make incorrect decisions with too much confidence. The researchers call for more “humble” AI systems: tools that reveal uncertainty and encourage additional information gathering when recommendations are uncertain.

This lesson extends beyond medicine.

Financial recommendations, education advising, legal analysis, HR decisions, and institutional risk assessments all require calibrated uncertainty.

AI governance translation:

Confidence calibration is a governance control.

Board Move:

Require high-impact AI systems to:

  • Disclose uncertainty
  • Identify evidence limits
  • Support challenge
  • Preserve human judgment before consequential action

8. Education Is Moving From Bans to Governance-by-Design

New America argued that blanket edtech bans miss the mark.

K-12 Dive reported that a Washington district expects to save up to $250,000 by using AI-supported “vibe coding” and agentic software engineering to replace some edtech subscriptions and build district-led tools.

This is exactly the governance tension in education.

  • Blanket bans-binary thinking- may be too blunt.
  • Ungoverned experimentation may be too risky.
  • The middle path is discernment and purposeful governance.

AI governance translation:

Education leaders need use-case governance, not panic bans or laissez-faire adoption.

Board Move:

Require an education AI use-case review that addresses:

  • Student data
  • Teacher autonomy
  • Accessibility
  • Procurement
  • Auditability
  • Instructional quality
  • Appeal pathways

9. AI-Generated Content Is Overwhelming Detection and Provenance Systems

Deezer reported that AI-generated tracks now make up 44% of all new music uploaded to its platform.

This is not only a music-industry story. It is a provenance story.

When AI-generated content becomes cheap, abundant, and scalable, institutions need mechanisms for:

  • Authenticity
  • Labeling
  • Detection
  • Rights management
  • Fraud monitoring
  • User trust

AI governance translation:

Content provenance is becoming institutional infrastructure.

Board Move:

Ask whether the institution can identify, label, manage, and respond to AI-generated content in communications, research, education, marketing, records, and customer-facing systems.

10. Insurance Markets Are Beginning to Price AI Uncertainty

The information reported that major insurers are moving to exclude AI-related damages from some corporate insurance policies, with state regulators approving many requests.

This may become one of the most important signals of board-level governance.

When insurers begin carving out AI risk, the institution cannot assume that “we are covered” remains true. AI governance now has to include risk financing.

AI governance translation:

AI risk is becoming insurable, excluded, priced, or contested—and boards need to know which applies.

Board Move:

Require legal, risk, procurement, and insurance teams to review whether AI-related harms are:

  • Covered
  • Excluded
  • Capped
  • Ambiguous
  • Unpriced
  • Dependent on vendor terms

The Seba Framework

The 12 Ps of Responsible AI Oversight ©

This week’s signals fit clearly into the full Seba 12 Ps framework:

  • Purpose — mission alignment versus convenience adoption
  • Problems — what decision problem is actually being solved
  • Profits — who benefits versus who bears risk
  • People — patient, student, worker, creator, and public impact
  • Planet — compute, energy, infrastructure, and scaling implications
  • Process — monitoring, updates, escalation, rollback, and incident learning
  • Policy — rules governing the use case and its limits
  • Protections — red lines, vulnerable groups, and complaint pathways
  • Privacy — data access, retention, exposure, and secondary use
  • Provenance — evidence, authorship, data lineage, model lineage, and traceability
  • Preparedness — leadership competence, governance cadence, and operational readiness
  • Product Ownership — who owns outcomes once AI shapes action

Six Ps Feel Especially Active This Week

Preparedness

Because boards and executives still disagree on who owns AI, while capability is moving into core operations.

Process

Because agentic tools now require access control, logging, monitoring, escalation, and rollback.

Privacy

Because education, the workplace, healthcare, and coding tools increasingly operate across sensitive data environments.

Provenance

Because AI-generated content, model competition, and vendor claims all require traceability and verification.

People

Because AI is shaping work, clinical judgment, social skills, student experience, creative markets, and institutional trust.

Product Ownership

Because insurance exclusions and vendor complexity make one thing clear: institutions cannot assume someone else will own the harm.

A Simple Example

If a bank, university, hospital, school district, or professional services firm deploys an AI agent that can:

  • Access internal systems
  • Produce work product
  • Coach staff
  • Process sensitive information
  • Influence decisions
  • Continue repeatable work over time

However, no one has documented its permissions, evidence base, limitations, insurance status, escalation path, or named owner, so several Ps fail at once.

  • Preparedness fails because leadership cannot explain the operating model.
  • Process fails because the workflow lacks controls.
  • Privacy fails because sensitive data may move through unclear channels.
  • Provenance fails because outputs cannot be traced or challenged.
  • People fail because users absorb the consequences.
  • Product Ownership fails because accountability diffuses at the moment it matters most.

That is how AI governance breaks in practice.

  • Not dramatically.
  • But by small decisions made without proper AI-informed frameworks and oversight.

Board-Ready Next Step

Require an AI Control & Coverage Sheet

If you do only one thing this quarter, require an AI Control & Coverage Sheet for every consequential AI use case.

  1. One page.
  2. Named owner.
  3. Board-reviewable.
  4. Updated after deployment.

At a minimum, it should answer ten questions:

1. What work is the AI system doing?

Drafting, coding, advising, triage, routing, scoring, coaching, monitoring, analysis, customer interaction, or execution?

2. What level of authority has been delegated?

Recommendation only, supervised action, partial execution, computer use, autonomous task completion, or ongoing repeatable work?

3. What systems and data can it access?

Public, internal, confidential, student, patient, employee, customer, financial, legal, privileged, or regulated data?

4. Who owns the use case?

Executive owner, operational owner, technical owner, legal/compliance owner, and escalation owner.

5. What evidence supports deployment in this context?

Vendor claims, independent validation, internal pilot results, sector-specific benchmarks, safety testing, red teaming, or post-deployment monitoring.

6. What are the known limitations?

Hallucination, overconfidence, bias, privacy exposure, cyber misuse, workflow mismatch, accessibility gaps, or domain limitations.

7. What gets logged?

Inputs, outputs, actions, overrides, access events, errors, escalations, user complaints, incidents, and near misses.

8. What triggers escalation or rollback?

Unsafe output, privacy incident, accuracy threshold breach, bias signal, cyber warning, user harm, vendor change, regulatory concern, or insurance exclusion.

9. What is covered or excluded by insurance?

AI-related errors, agentic actions, data exposure, discrimination claims, IP disputes, professional liability, cyber events, or third-party harms.

10. Who can pause or stop use?

Name the role, not the committee.

That sheet turns “we are using AI” into:

“We know what AI is doing, what it can touch, what evidence supports it, what risk remains, what coverage exists, and who owns the outcome.”

AI Governance Book Update

As my forthcoming AI governance book moves closer to release, this week’s signals reinforced one of the book’s central claims:

AI governance is not a principles exercise. It is an ownership exercise.

The hardest questions are no longer abstract:

  • Who owns the AI strategy?
  • Who owns AI risk?
  • Who owns AI-enabled work?
  • Who owns AI-generated harm?
  • Who owns vendor failure?
  • Who owns the gap between what the tool can do and what the institution is prepared to govern?

A practical governance before consequential AI becomes normalized inside institutions that were never designed to absorb this much delegated capability at this speed. More soon.

Podcast Note

This week’s issue also connects directly to the broader arc of AI Governance with Dr. Freddie Seba.

Across recent episodes, the same theme keeps resurfacing:

  • AI governance is no longer only about whether technology is ethical in the abstract.
  • It is about whether leaders can create the operating conditions for responsible use.

That means:

  • Governance before scale
  • Evidence before trust
  • Control before autonomy
  • Ownership before harm

And in sectors like education, healthcare, financial services, public leadership, and youth-facing technology, those questions are not optional. They are fiduciary.

Final Thought

AI Governance Is Entering Its Control Phase

The institutions that lead will not be the ones with:

  • The most AI pilots
  • The most vendor demos
  • The most ambitious adoption language
  • The most polished AI strategy deck

They will be the ones who can answer:

  • What is the AI doing?
  • Who owns it?
  • What can it access?
  • What evidence supports it?
  • What happens when it fails?
  • What is excluded from coverage?
  • Who can stop it?

That is the work now. Not AI enthusiasm. It is AI governance. And increasingly, that is the difference between innovation maturity and governance debt.

About the Author

Dr. Freddie Seba is a global executive, scholar-operator, author, and Silicon Valley–based innovation leader whose work sits at the intersection of AI governance, digital health, and mission-driven institutional leadership.

With more than 20 years of experience spanning San Francisco, Silicon Valley, and the global stage — across banking, fintech, digital health, startups, and higher education — he brings together the humanistic lens of institutional leadership, the strategic discipline of business execution, and the policy rigor of global institutions.

He holds an EdD in Organizational Leadership, an MBA from Yale, and an MA in International Policy from Stanford. He writes and speaks on AI ethics and AI governance for leaders, boards, and trustees across fintech, health tech, healthcare, higher education, and related institutional contexts. His work translates fast-moving AI developments into plain-language governance for senior decision-makers: decision rights, risk tiering, vendor accountability, monitoring, and incident preparedness.

Global Executive | Silicon Valley Founder | AI Governance Scholar-Operator | MBA, Yale • MA, Stanford • EdD, USF

Gratitude

Grateful to the researchers, practitioners, technologists, journalists, policy thinkers, educators, clinicians, and institutional leaders whose work shaped this week’s analysis.

This issue draws on reporting and research from TechCrunch, Fortune, Stanford HAI, MIT News, Anthropic, OpenAI, K-12 Dive, New America, Deezer, the Ada Lovelace Institute, the European Commission, BMJ Health & Care Informatics, ScienceDirect, The Information, and others working to make AI governance more concrete, evidence-based, and operational.

Transparency + Disclaimer

Educational content only. This newsletter does not constitute legal, medical, clinical, insurance, financial, or professional advice.

Drafted and refined with AI-assisted tools for synthesis and clarity. Final editorial control and responsibility remain with the author.

© 2026 Freddie Seba. All rights reserved.

Selected Sources Reviewed This Week

Hashtags

#AIGovernance #ResponsibleAI #BoardOversight #AILeadership #AIEthics #RiskManagement #HealthcareAI #FinancialServices #EdTech #Cybersecurity #AgenticAI #AIInsurance #TrustInfrastructure #DigitalHealth #HigherEducation #FiduciaryGovernance