OpenAI

Incident Summary

Valehart is tracking a broader set of reported OpenAI billing, purchasing, credit-consumption and organisation-control failures rather than a single customer-support dispute.

The investigation began following an August 2026 report in which an OpenAI user described three consecutive USD $500 charges associated with an organisation they stated they did not recognise and could not access, followed by a fourth attempted USD $500 charge.

Subsequent public reports describe additional financially consequential or poorly auditable behaviour, including credits decreasing without corresponding visible use, immediate charging when enabling auto-reload, repeated purchases where the user states auto top-up was never enabled, and rapid repeated auto-reloads.

These reports do not establish that every observed behaviour shares one technical root cause. Collectively, however, they support an ongoing investigation into purchasing authority, credit accounting, transaction visibility, auto-reload behaviour, organisation association and the ability of users or organisations to audit and constrain financial consumption.

Continuing Public Reports

Evidence Tracks

1. Financial / Purchase Controls

• Public reports include repeated transactions ranging from individual unexpected purchases to dozens of charges over short periods.
• Reported financial consequences include hundreds to more than USD $1,500 in transactions.
• At least one report alleges repeated purchases despite auto top-up never having been enabled.
• Another report alleges that enabling auto-reload itself caused an immediate USD $100 purchase without a separately disclosed purchase step.
• The evidence supports continuing concern about the controls governing when paid resources may be purchased or consumed.

2. Credit Metering & Auditability

• Users have reported credits or usage allowances decreasing without being able to reconcile the reduction against their visible activity.
• The absence of transaction-level or task-level visibility can prevent users from determining which operation consumed a purchased resource or allowance.
• This creates a distinct accounting and auditability concern even where no separate payment-card transaction occurs.

3. Organisation Association

• The triggering investigation involved charges associated with an organisation the affected user stated they did not recognise and could not access.
• Valehart separately demonstrated that an OpenAI workflow accepted an organisation ID belonging to a different independently controlled OpenAI account.
• Access and promotional benefits were subsequently provisioned using the supplied organisation identity.
• This establishes cross-account organisation association in the tested workflow.
• It does not establish that the demonstrated workflow caused the reported billing incidents or that payment methods can be transferred using the same mechanism.

4. Abuse / Resource-Leeching Risk

• Where purchasing or credit consumption occurs without adequate authorisation, visibility or account isolation, users can face direct financial loss or depletion of purchased resources.
• Comparable weaknesses affecting organisation-associated resources could expose organisational budgets, credits or purchasing capacity to unauthorised consumption.
• Cross-account organisation association increases the relevance of this risk, although a causal link between the demonstrated association behaviour and reported financial transactions has not yet been established.

5. Vulnerability Disclosure Handling

• 07-08-2026 — Valehart began attempting to route the billing/organisation-control concern to OpenAI while documenting related reports.
• 09-08-2026 — OpenAI personnel were informed that Valehart's concern extended beyond the triggering user report and that independent testing and a proof of concept were available.
• 11-08-2026 — Further direct and public disclosure attempts summarised organisation, billing and visibility concerns.
• 01-09-2026 — Valehart again raised the unresolved matter with additional public evidence and organisation-security findings.
• As of 03-09-2026, no technical acknowledgement, vulnerability intake, root-cause explanation or request for the available proof of concept has been received.

Current Assessment

Incident status: ACTIVE / UNRESOLVED. Public reports throughout August 2026 continue to describe financially consequential billing, purchasing, credit-depletion and usage-accounting behaviour.

Financial exposure: ONGOING. Reported cases demonstrate the potential for rapid depletion of paid resources and repeated financial transactions, with consequences ranging from individual consumers to organisation-associated accounts and budgets.

Control risk: UNRESOLVED. The evidence raises continuing questions concerning purchasing authority, auto-reload behaviour, credit accounting, transaction visibility and account/organisation isolation.

Technical mechanism: UNRESOLVED. The available evidence does not establish a single root cause connecting every reported case. Valehart's independent cross-account organisation test establishes a separate control weakness relevant to the investigation but does not establish that it caused the reported financial transactions.

Disclosure: UNRESOLVED. Valehart has offered technical findings and an available proof of concept without receiving technical engagement as of 03-09-2026.