Apple’s reported vulnerability-report cap puts AI security triage under scrutiny

Apple’s reported decision to limit how many vulnerability reports a researcher can keep open has developed into a wider dispute over how security teams should handle growing volumes of AI-generated submissions.

The policy was reportedly introduced in June 2026 and surfaced through Financial Times reporting in early August. A September 5 column from 9to5Mac renewed criticism of the measure, but did not contain a separate Apple announcement or establish a new programme change. The issue is therefore an ongoing disclosure-policy question rather than a newly imposed restriction.

The Financial Times reported that Apple had limited the number of security reports a researcher could have open as volumes of AI-generated submissions grew. The newspaper also described a 30-day cool-off after the threshold is reached. Apple said researchers may request a higher quota at any time, so the measure is not an unconditional block on further reporting.

The number of reports permitted before the restriction applies has not been disclosed. Apple’s public bounty guidelines also do not describe the 30-day mechanism, the criteria for approving quota increases, how quickly requests are reviewed or whether researchers have a separate route for urgent submissions.

AI volume and report quality

Apple has linked the intake pressure to increasing volumes of AI-generated security reports. Its official guidance says submissions must be complete and actionable, with a reliable reproduction method or proof of concept. Researchers are advised to avoid lengthy descriptions generated by AI and must validate their findings themselves.

The company classifies theoretical issues, including vulnerabilities identified by AI without proper human validation, as ineligible. Apple says unvalidated reports produced by large language models can consume investigation capacity and delay work on serious security issues.

The guidance draws an important distinction between AI-assisted vulnerability research and automatically generated claims that a researcher has not confirmed. For AI-related findings, a human researcher must establish that the reported weakness is reproducible and actionable before submitting it.

Apple already has a separate sanction for researchers who repeatedly file ineligible reports. Its guidelines state that this behaviour can result in a 180-day processing pause. That rule is distinct from the 30-day cool-off reported by the Financial Times and should not be treated as another description of the same restriction.

Target Flags provide another quality check

Apple’s Target Flags offer another mechanism for separating demonstrable vulnerabilities from theoretical findings. Announced as part of the company’s bounty-programme changes in October 2025, the flags are intended to provide objective evidence of exploitability.

After Apple verifies a Target Flag, an eligible report can qualify for an accelerated award. The mechanism gives researchers a structured way to demonstrate exploitability, although it does not remove the need for expert review or explain how Apple prioritises valid submissions once a researcher has reached the open-report cap.

The policy tension is that a volume control does not necessarily distinguish between automated noise and a legitimate run of multiple findings. A researcher using AI as part of a human-led, validated process could still produce reports at a rate that reaches an undisclosed limit. Conversely, unrestricted intake of unverified model output could divert triage resources from reports that already contain reproducible evidence.

Bradley Chambers of 9to5Mac argued that Apple should address the problem through improved triage rather than submission caps. That is an editorial position, not an independently established assessment of the policy’s effectiveness. Apple has not published submission volumes, false-positive rates, backlog figures or triage-time data that would allow the effects of the cap to be measured publicly.

Bynario account remains unverified

The Financial Times used the experience of seven-person security company Bynario to illustrate the possible consequences for high-volume researchers. Bynario reportedly said it encountered the restriction after submitting eight reports in 2025 and five in 2026, while it had identified further possible macOS issues.

Bynario also claimed, according to the Financial Times, that those findings included a privilege-escalation chain that could potentially provide full control of a Mac. The claim has not been independently verified through an Apple advisory, CVE record or public technical disclosure. No affected versions, technical cause, exploitation evidence, disposition or fix status have been established for the unidentified chain.

The Financial Times also reported Apple’s statement that it was in contact with Bynario and reviewing the company’s additional findings. Publicly documented Bynario research must not be assumed to describe the same vulnerabilities covered by that account.

The policy story itself does not establish that an Apple product has been compromised, that data was accessed or that customers face a specific vulnerability requiring mitigation. There is no applicable patch or end-user security action. Researchers can request a higher quota and follow Apple’s requirements for concise, reproducible and human-validated evidence, including Target Flags where applicable.

For Australian researchers and organisations using Apple products, the relevance is indirect. The concern is how quickly legitimate findings reach Apple’s security teams, not a documented Australia-specific incident, regulatory action or policy response.

The Hack Academy has previously covered Apple’s bounty expansion, AI-assisted Apple vulnerability research and accelerated security updates. The report-intake limit presents a separate question about how Apple manages access to its disclosure pipeline while trying to reduce the burden of unvalidated output.

Key details remained unavailable at the evidence cutoff on 11 September 2026. Apple had not publicly specified the exact quota, the response time for an increase request, the current status of the reported June policy or metrics showing whether it had reduced triage delays. The disposition of Bynario’s unidentified findings also remained unresolved.

Reproducible evidence gives reviewers a firmer basis for assessing a tentative vulnerability finding. If you want to develop the surrounding technical foundations, consider The Hack Academy’s online cybersecurity training programme, which offers self-paced learning and practical exercises alongside theory and lists internet vulnerabilities and penetration testing among its subjects; keep all testing authorised and controlled.