Personal and work AI need an identity boundary

Map identities and contexts before they converge

Inventory personal accounts, managed work accounts, tenant and workspace memberships, developer organizations and projects, API keys and service accounts, Codex environments, connected applications, enterprise identity providers, devices, billing relationships, and recovery channels. For each, record owner, administrator, authentication, role, data region, contract, policy, allowed information, approved use, action rights, monitoring, retention, export, deletion, and exit path. Define whether a person may link or switch identities, move chats or files, invoke company connectors from personal space, use personal context in work, or use work results outside the tenant. A familiar interface should not obscure which legal and security boundary governs the active context.

Decide what memory and data may cross

Classify prompts, conversations, memories, custom instructions, uploaded files, repository contents, code and artifacts, connector data, browsing, voice or image records, outputs, feedback, evaluations, and logs. State which context is personal, company, client, regulated, confidential, privileged, export-controlled, or public; who can retrieve it; whether it informs later responses; and how the user sees and corrects it. Prevent automatic context transfer unless the purpose, source, recipient, retention, and user notice are authorized. Keep data access distinct from permission to disclose, summarize, train, publish, commit, send, purchase, or execute.

Design portability, monitoring, and offboarding together

Set rules for employee-owned prompts and methods, company records, client deliverables, generated code, reusable agents, credentials, and business continuity. Monitoring should be purpose-limited, disclosed, role-restricted, and tested so personal activity is not swept into workplace review. At role change or exit, revoke enterprise sessions, connectors, keys, delegated agents, repository and data access, shared links, scheduled work, and recovery paths; transfer company-owned artifacts through an approved process; preserve required records; and verify that personal access no longer reaches them. Do not rely on removing a single seat when identities and integrations span several products.

Test boundary failures before expanding

Use scenarios in which an employee switches accounts, pastes client data into a personal chat, invokes a work connector from the wrong context, uses personal memory in a regulated answer, exports a reusable agent, shares a link, changes roles, leaves the company, and later tries recovery. Confirm that identity and policy are visible, restricted actions fail, evidence is retained appropriately, company artifacts remain available, personal material stays separate, and revocation closes every route. OpenAI's article supports the provider's convergence thesis and reported scale. It does not establish a buyer's governance, security, privacy, labor, records, IP, continuity, or business outcome.

Turn this source into a reviewable decision

For AI for CEOs, use this briefing as a dated decision record rather than a substitute for the source. Preserve The Work Now Within Reach, the exact URL, the September 9, 2026 review date, the supported facts above, the editorial interpretation, the limitations, and any buyer-specific evidence. Link that record to the decisions most directly affected: Operating-model redesign; Board governance and oversight; Enterprise resilience and risk; Leadership capability and decision practice. State whether the source changes the scope, evidence requirement, control, sequence, or only the language used to describe the decision.

Before action, name the accountable owner, affected population and workflow, exact offering or configuration, source data and rights, human decision point, exception and appeal path, complete cost, expected benefit, failure and stop conditions, retained evidence, and next review date. Keep official facts, provider statements, buyer observations, representative tests, measured outcomes, editorial inferences, and unknowns visibly separate. Reopen the record when the source, offer, model, integration, data, policy, population, responsible person, or measured result changes.

Limitations and unknowns

This briefing uses OpenAI's official article whose RSS timestamp precedes the September 8, 2026 cutoff; it is not claimed as a post-cutoff event. User and business counts are OpenAI-reported, and convergence is provider framing rather than evidence of a buyer's architecture or outcome. Product boundaries, contracts, identity features, memory, retention, data use, connectors, export, and deletion can change. Current agreements and settings, identity and access tests, data maps, exit procedures, and qualified security, privacy, records, IP, employment, procurement, regulatory, and legal review control.