FAQ: Maintaining test security in the age of technology

The following table provides a summary of standards and guidelines on test security that may be relevant.

At the core of test security is the need to preserve the integrity and intended interpretation of test scores. That integrity depends not only on preventing prior access to test materials or items, but also on protecting secure content, maintaining standardized administration conditions, limiting unauthorized assistance or use of prohibited tools, and preventing behaviors that introduce construct-irrelevant variance. Therefore, psychologists are encouraged to become familiar with the various ways test security can be compromised, including cheating, unauthorized disclosure or copying of test content, third-party assistance, and technology-enabled misconduct, because these threats can reduce the validity, fairness, and usefulness of test scores for decision-making purposes.

Psychologist responsibility may include, depending on the assessment context, ensuring that everyone involved in the delivery of an assessment has awareness of the importance of test security, including ensuring that test takers don’t have access to unauthorized materials and resources (e.g., notes, internet access, access to AI agents or similar), protecting the assessment content from unauthorized users, using encryption or passcodes when storing data, and having a process for determining how any potential security breaches should be reported, both to their clients (or test takers) as well as to the test publisher(s), etc. It is important to train all support staff on test security and how to address problems without viewing an individual’s work, or if that is not possible, to require staff to sign a confidentiality agreement to not disclose test content.

Programs should prevent test takers from capturing and sharing test material. Browser lock technologies can prevent screenshots, unauthorized access to apps and web sites, and the ability to download software, install agents, etc. Consider providing individual ID codes or following protocols that allow test takers access to only the parts of the test that are relevant to them. If testing in a group setting, privacy screens are recommended.

Proctors may be present to monitor the assessment session and may be asked to sign a non-disclosure agreement protecting the confidentiality of any material that is inadvertently observed while assisting a test taker.

In employment settings, the issue of maintaining test security is much more complex because there is a growing interest in allowing potential employees to be able to complete assessments on their own mobile devices (Tippins, 2015). Therefore, the ability to disable features or implement some of the practices described above may not apply. In the case of unproctored internet-based testing (UIT), a test taker has responsibility for ensuring no third-party observation, recording of the assessment in any form, or use of prohibited tools, including AI, when such tools would compromise the intended meaning of the score. Test takers should be informed of the consequences for violating these rules.

As new technologies emerge, these tools warrant a periodic review and update to program security procedures and test-taker instructions to address evolving threats.

Generative AI and related tools can introduce new test-security breaches and construct-irrelevant variance and, depending on the context, may warrant communications with test takers, in the same manner as other unauthorized aids (e.g., calculators or dictionaries), third-party assistance, and unauthorized recording or copying of test materials.

In practice, test takers may attempt to use generative AI and connected devices to obtain real-time assistance or to covertly communicate during an exam. For example, they may consult AI chatbots to generate or rewrite responses, use voice assistants with discreet audio prompts, or rely on messaging apps to share questions with someone offsite. Wearables can add another channel for concealment, including smartwatches used for short text exchanges, notifications, or stored notes, and wireless earbuds or similar to receive spoken cues. Some may also use small cameras to capture test items for later reconstruction and sharing. Smart glasses represent another way that test takers might be able to leverage technology to cheat by providing outside help, access to unauthorized materials, or allowing for the unauthorized copying of secure content. Training test administrators to recognize wearables capable of supporting cheating can be a critical safeguard against security risks that could compromise score validity. These tactics can undermine score validity by introducing outside help and unauthorized copying of secure content.

As part of prevention and deterrence, testing programs are encouraged to define (a) what constitutes prohibited AI-assisted responding (e.g., generating, rewriting, or coaching answers during the assessment), (b) what constitutes prohibited post-test behavior (e.g., reconstructing or sharing test content), and (c) any permitted supports (e.g., approved accessibility accommodations) in ways that are consistent with the intended interpretation and use of scores. When feasible, test takers may be asked to acknowledge these expectations in advance (e.g., non-disclosure agreement, security attestation) and be informed of the consequences associated with violations.

In addition, programs using remote formats may consider layered protections appropriate to the stakes and construct (e.g., strengthened identity verification, increased monitoring with AI, verification testing or follow-up assessment where justified, and test-design strategies that limit content exposure). Detection indicators are best treated as evidence warranting review rather than definitive proof of misconduct, with policies establishing procedures for investigation, documentation, and consistent treatment of test takers.

Although organizations may vary in their security policies, it is recommended that psychologists do what they can to ensure that attention is paid to test security, that those associated with test administration (e.g., on site administration personnel, IT personnel) have awareness of their individual obligations regarding securing materials, that monitoring for security breaches is appropriate, and that consequences for security breaches are clear to all involved (i.e., test administrators, test takers).

Cheating is associated with ability and knowledge tests (that are characterized by correct and incorrect responses) and entails the use of illicit aids to obtain and produce the correct answers to test items or any efforts to misappropriate test content. Illicit aids may include both traditional resources (e.g., unauthorized devices or notes) and technology-enabled resources (e.g., unauthorized online resources or AI-enabled tools) when their use results in responses that do not reflect the test taker’s own standing on the constructs of interest. Cheating poses a measurement threat and concern because it is a major source of psychologically based construct-irrelevant variance. Cheating behaviors may produce test scores that do not accurately reflect the test taker's standing on the constructs of interest, thereby undermining the validity and interpretability of the test.

Approaches to deter cheating can be conceptualized as (a) to discourage cheating attempts, and (b) to make it more difficult to engage in cheating. Approaches to discourage cheating include monitoring, the saliency of detection techniques in general, and warnings and threats. In the case of unproctored internet-based testing (UIT) or remotely delivered assessments, monitoring is often electronic in nature (e.g., identity verification, session monitoring by humans and AI, browser lock technologies, room scans, keyboard monitoring, etc.). However, in proctored settings, the use of human observation and supervision is the primary means to curtail, prevent, and detect cheating. To deter individuals from cheating in both contexts, programs may inform test takers about test-security expectations (including prohibitions on unauthorized aids such as AI-enabled tools), the presence of detection techniques and data forensics, and the consequences of violations; when feasible, this may include a pre-test acknowledgement of these expectations.

Deterring cheating by making it more difficult focuses primarily on test design characteristics and features such as the use of multiple test forms, computer adaptive tests (and similar dynamic types of delivery), and speeded tests. The use of multiple test forms and computer adaptive tests seeks to address concerns arising from the pre-knowledge of test items. Speeded tests have been found to reduce cheating on ability assessments by making it harder for individuals to use unauthorized aids. Yet, with recent technological advances such as agents that run silently in the background and use computer vision to instantly read and answer questions, it's become much easier to access unauthorized materials within milliseconds. These methods make cheating nearly impossible to detect when relying solely on timing as an indicator.

Occasionally, a proctor directly observes and “catches” a test taker cheating; increasingly, AI is being used to augment the proctor’s ability to identify cheating behavior. In these cases, AI supports the human proctor, and the human makes the final decision on what action, if any, to take.

Statistical detection methods, commonly known as data forensics, can be used to flag instances of anomalous test responses. For instance, (a) the similarity of (pairs of) scores or response patterns (of errors and correct responses) or (b) deviations from some known or expected distribution of scores which may be probabilistic or actual (retest) scores are just two examples of statistical methods to detect anomalies in test responses.

Other methods include the use of biometric identification systems such as fingerprint, iris, and retina scans to verify the identity of the test taker, and webcams to monitor test takers during the test administration. Other layers of technological security include authentication via keystroke analytics, electronic monitoring, and controls such as real-time data forensics, browser lockdown, and operating system and internet access controls. It is important to emphasize that statistical detection methods do not actually detect cheating; they only provide evidence or data that permit inferences about the presence of cheating.

Finally, some programs use items that have been exposed, released, or inappropriately obtained with minor changes that are unnoticeable to test takers but that can be used to identify pre-knowledge. These items have been called Trojan Horse items. Differential performance on these items versus the other items on a test can help detect whether the examinee obtained access to the items inappropriately.

Because any single indicator can yield false positives, testing programs are encouraged to specify in advance how indicators will be reviewed and what follow-up steps (e.g., investigation, documentation, retesting) will be used before adverse decisions are made.

For the purposes of this FAQ, test security breach involves the unauthorized access, disclosure, copying, distribution, or use of secure test content (or other actions that compromise the confidentiality of materials or the integrity of scores). Examples include capturing or sharing test items, using unauthorized aids (including prohibited technologies) in a way that changes what the score means, permitting unapproved third‑party assistance, or mishandling secure materials such that others could obtain them.

An administration irregularity is a deviation from standardized administration conditions that may affect fairness, comparability, or score interpretation but does not necessarily involve exposure or theft of secure content. Examples include timing errors, interruptions, technology failures, accommodation delivery errors, or proctoring lapses (e.g., inadequate monitoring) that create nonstandard conditions. Programs are encouraged to define these categories in advance, use consistent decision rules for classification, and document how each type of event will be investigated and how scores will be used, invalidated, or retested as appropriate.

“Secure” test materials are those that require restricted access because public exposure would undermine the validity, fairness, or utility of the assessment and/or violate copyright or test-use agreements. Depending on the program and test, secure materials commonly include: (a) live item content (stems, options, prompts, passages, graphics), (b) complete test forms and assembly rules, (c) administration and examiner manuals when they include verbatim items, prompts, discontinue/reversal rules, or other content that enables reconstruction of the assessment, (d) scoring keys, rubrics, anchor responses, scoring algorithms, and calibration materials, (e) item bank exports, locally cached test content, and screenshots/recordings of test screens, (f) credentials and access mechanisms (proctor passwords, tokens, admin URLs or settings that provide elevated access), and (g) item- and form-level analytics that could enable item identification or reconstruction (e.g., detailed item statistics, exposure rates, and flagged-item lists) when shared outside authorized roles.

Maintain documentation sufficient to (a) support fair, consistent decisions for test takers, (b) protect test content, and (c) meet legal, privacy, and contractual obligations. Common elements include:

• Incident report : date/time, location or delivery mode, assessment name/form (if applicable), session identifiers, proctor(s) and key staff involved, and a concise narrative of what was observed or flagged.
• Classification and rationale : whether the event is treated as an irregularity or breach (and why), and what policies/decision rules were applied.
• Evidence and artifacts (as permitted): proctor statements, screenshots/photos where allowed, confiscated materials logs, remote-proctoring flags and notes, system audit logs, identity verification results, and any communications with the test taker.
• Test-taker notice and response : what was communicated to the test taker, their explanation (if obtained), and timelines for response.
• Investigation actions : who reviewed the case, what data were examined, and any consultation with the publisher/vendor, legal counsel, institutional officials, or ethics/compliance resources.
• Outcome and follow-up : score decision (e.g., accepted, delayed, invalidated), retest decision and conditions, sanctions if applicable, and any test-form remediation (e.g., retiring exposed items/forms).
• Chain of custody and access controls : who had access to evidence, where it was stored, and when it was transferred or destroyed.
• Retention and disposal record : how long records will be kept and when/how they were securely destroyed, consistent with organizational retention schedules and applicable laws/contracts.

Recording and log handling considerations may vary by legal jurisdiction, platform, assessment context, retention requirements, privacy obligations, contractual terms, and the purpose for which the recordings or logs are collected or reviewed. Remote-proctoring recordings and logs (e.g., video, audio, screen capture, chat transcripts, identity verification artifacts, event flags, and system audit logs) may warrant treatment as sensitive assessment records, depending on the context. Depending on the context, programs may consider collecting and retaining only what is necessary for security and score use, providing clear notice to test takers about what is recorded and how it will be used, and implementing role-based access so that only authorized personnel can view recordings or detailed logs.

Retention periods may be defined in advance (often longer when stakes are higher or when appeals windows are longer) and aligned with privacy requirements, legal and regulatory compliance, contractual terms, and the time needed to investigate anomalies. Access may be logged, and reviewers trained to avoid unnecessary exposure to secure content when reviewing flagged sessions. At the end of the retention period (or once a case is closed, if sooner is allowed), recordings and logs may be securely deleted, including from vendor dashboards, exports, and any local copies, and the program may be required to retain documentation that the destruction occurred.

Programs should strive to establish an appeals process in advance that is transparent, timely, and applied consistently. When a score is questioned due to potential irregularities or breaches, the program may document the basis for concern, allow the test taker an opportunity to respond, and have qualified personnel review relevant evidence (e.g., proctor reports, system logs, data-forensics indicators, and recordings where used) using predefined decision rules.

At the same time, appeals procedures should strive to protect secure test content and third-party privacy; for example, programs may summarize findings without releasing items, scoring keys, or other protected materials. Possible remedies may be specified ahead of time (e.g., score hold, invalidation, supervised retest, verification testing, or reinstatement when concerns are resolved). Where consequences are imposed, documentation procedures strive to reflect the rationale, the evidence considered, and the path for further review consistent with legal and contractual requirements.

Training is role-specific and covers both standardized administration and test security. Training commonly includes: (a) what materials are secure and how they are to be stored, transmitted, and disposed; (b) permitted vs. prohibited aids (including technology and AI-related rules) and how to communicate these rules to test takers; (c) identity verification and check-in procedures; (d) how to monitor sessions (in-person or remote) while minimizing unnecessary exposure to content; (e) how to respond to suspected irregularities/breaches (pause/stop rules, documentation, escalation paths, and evidence preservation); and (f) confidentiality obligations, including any non-disclosure agreements and applicable legal/privacy requirements.

For IT and platform support, training may also include role-based permissions, secure configuration (e.g., account provisioning, logging, patching, device controls), incident triage that avoids viewing test content when possible, and the proper handling of recordings/logs. Refresher training is typically provided at onboarding, at least annually, and whenever policies, platforms, or threat conditions change (e.g., new remote-proctoring tools, new AI-related risks, or updates from publishers/vendors). Brief pre-administration reminders or checklists can reinforce key steps before each test window.

Testing programs strive to have a plan for how security issues will be handled before any occur. The plan may include how to investigate any violations and define the actions that will be taken based on the type and severity of the offense.

If the security breach results from the act of a psychologist, per the APA Ethics Code (American Psychological Association, 2017), the first step is to informally notify the psychologist of the violation. In some cases, the psychologist may be unaware of the security breach or may not have understood the necessity of test security and why the behavior is problematic. Such efforts would involve a request to stop the behaviors that contribute to the security lapse and correct the situation as appropriate. Persistent violations may require additional steps, such as contacting state licensing boards to file a formal complaint, contacting appropriate institutional authorities, and contacting test publishers regarding potential test use agreement violations or copyright violations; all actions should be taken to ensure there are no violations of confidentiality rights. In some cases, state laws and/or state board policies, administrative rules, or directives exist regarding the security of testing materials for student testing. If such laws, policies, or rules are pertinent to the security violation in question, there may be a specific prescribed procedure to follow. For educational testing, states may have investigative procedures, hearings, and specific sanctions (e.g., reprimands, job suspensions or dismissals, fines, imprisonment) for breaches of test security. These may or may not be applicable to psychologists and test security breaches, depending on the testing setting.

Test-security violations by other professionals (e.g., teachers in education settings, lawyers in court proceedings, physicians in health care contexts, employers in hiring contexts) may also be first addressed through informal, direct communication regarding the nature of the violation and why it is a concern. Here, too, the professional may be educated through this informal contact about the existence of a security lapse and about why test security is a concern, and a request can be made to remedy the breach if possible and/or to cease and desist any behaviors that are creating security violations. Persistent violations by other professionals may be addressed by contacting relevant ethical and professional societies and legal and oversight authorities (e.g., state boards, test publishers) when applicable (e.g., copyright restrictions, contractual obligations).

In some types of assessments, test administrators should inform test takers in advance about what may constitute a violation of test security (including unauthorized use of technology or AI-enabled tools, when applicable) and the consequences associated with any violation. Written policies provide control for guiding how test administrators respond to suspected security violations by test takers. If a security breach occurs during testing, the ITC Guidelines (2014) suggest pausing or stopping testing and explaining the reason for this to the test taker, confiscating any inappropriate material, and reporting the irregularity in testing.

Policies may address the following: investigative procedures; procedures for dealing with scores obtained under potentially problematic conditions including any retesting; procedures regarding the acceptable continued use of test materials after a breach; appropriate documentation and preservation of relevant evidence consistent with privacy, legal, and contractual requirements; reporting to a licensing board; and any civil or criminal legal action that might be taken against a test taker. In general, ensure consistency in treatment of violators, preserve the integrity of the testing process, and deters future violations. An appeals process is recommended.

When a new version of a test is published, psychologists are faced with the problem of what to do with the older version. Old versions of the test may be destroyed in a secure manner, in much the same manner that would occur with patient records. Paper materials can be recycled through a secure shredding process, or they can be shipped back to test publishers for secure recycling. In many cases, this latter service is free, and test publishers will sometimes send pre-paid postage labels. Selling or sharing old test materials with unqualified individuals may compromise test security and violate test-use agreements.

Outdated or unwanted electronic test materials (e.g., item banks, forms, administration manuals, scoring keys, local cached test content, exported reports containing items, and backups) may be retired and disposed of in a secure manner consistent with confidentiality, copyright, and contractual obligations. Practically, this means removing access (deprovisioning accounts/roles; revoking sharing links), securely deleting files from active systems and endpoints, and ensuring copies are also addressed (downloads, email attachments, shared drives, removable media, and cloud backups) according to the organization’s retention and secure-destruction procedures. When tests are publisher-provided or vendor-hosted, follow the vendor’s formal retirement/decommissioning process (including written confirmation, where feasible), if applicable, and avoid selling, sharing, or transferring electronic materials to unqualified users, as doing so may violate ethical responsibilities and test-use agreements.

Proper disposal of assessment instruments is an important consideration when psychologists retire or otherwise cease providing professional services (e.g., career transition, license revocation, serious illness, or death).

The disposal plan may be written, filed appropriately, shared with the designated responsible person, and include safeguards against breaches that would affect the integrity of test materials. Appropriate disposal may include donating items to a professional psychology training program, to another practitioner, or if dated, destroying the items in a manner that is consistent with those discussed in response to the preceding question.

Reproducing or distributing copyrighted material without authorization is illegal and can result in claims for civil damages, as well as possible criminal prosecution.

The reproduction and inclusion of copyrighted test items in scholarly and scientific publications and presentations might not only be a copyright violation but may also undermine test security by potentially placing test items in the public domain, making test takers’ pre-knowledge of test items a major threat to the utility of an instrument. Indeed, this concern applies to all tests whose validity would be compromised if test takers had pre-knowledge and could thus “practice” the items. Instead, presenters may use sample items provided by the test company, or items that have previously appeared in research studies as sample prompts or items. This allows for the inclusion of detailed examples that are consistent with the instrument but that do not reflect actual content.

Third-party observation can occur in the form of another person being present in the testing room with the examiner and examinee or in the form of a digital (audio or visual) recording of an evaluation. Much has been written regarding the impact of third-party observers on test results from psychological or neuropsychological evaluations, but most of these articles focus primarily on the issues of validity of test scores, violation of standardized test administration, and impact on the test taker rather than the threat to test security by the presence of third-party observers. Breaches of test security due to possible public disclosure resulting from third-party observations should be considered in light of copyright law, in addition to the ethical obligations of psychologists to protect the integrity of secured tests.

Creation of a retrievable record resulting from third-party observation also introduces the risk for test security breaches. However, some courts routinely mandate that attorneys be allowed to be present for their clients’ evaluations or that the evaluation be recorded, creating retrievable records. Potential remedies to these test security threats include but are not limited to refusing to perform the evaluation, restricting the presence of third-party observation to the interview portion of the evaluation, specifying the limits of the role of the observer, discussing the subsequent use of the records created prior to the evaluation, and seeking protection orders from judges mandating test security protections.

To reduce the risk of inadvertent disclosure of secure test content, test materials (e.g., tests, administration manuals, scoring guides) are best maintained separately from patient records, and test publishers can facilitate this process by creating individual record forms (e.g., answer sheets) that do not include test content. Psychologists involved in litigation have an obligation to make available the records they used in rendering expert opinions. The legal obligation imposed by a court order or subpoena to release test data, including content, can present a dilemma for psychologists confronted with the possible contradictions between following the law and fulfilling their ethical obligation to protect test content.

Often, psychologists are asked to disclose test materials to lawyers and other individuals who are neither trained in psychological testing nor ethically or legally obligated to keep these materials confidential. Sometimes these materials will be admitted into evidence as part of the public record. These actions can potentially erode the reliability and validity of the instruments as the test items become more widely available to anyone trying to obtain access to them.

In response to this, psychologists’ practices vary widely with some individuals automatically releasing raw data in response to a subpoena and others refusing to do so under any circumstances. One of the best ways psychologists can minimize threats to test security in forensic evaluations is to request that the materials only be released directly to a qualified psychologist who is also bound by the same duties to protect test security. Citing relevant state laws and position papers can be useful in facilitating such secure disclosure. Alternatively (or in addition), a protective order from the judge can be obtained by the psychologist, obligating all parties to maintain test security and document all plans for disposal at the conclusion of the case. This protective order can specify the reasons for limited disclosure, the restrictions of unqualified individuals’ use of the data, and disposal of the data at the conclusion of the case. Restrictions may include prohibiting parties from making copies of the materials, requiring that all materials be returned to the psychologist at the close of litigation, and ordering that the record be sealed if test content is admitted as part of the public record.

A braindump site is a site that posts, sells, or facilitates sharing of actual or reconstructed test questions in an online forum or website. The posting of test content and material online present a persistent challenge to test security because operators and contributors may be anonymous, located outside the publisher’s jurisdiction, or ablet o repost on other sites after removal. Although rights holders may request removal or disabling of access to infringing test content through available takedown or notice processes, enforcement can be difficult when the site, host, or operator is outside the relevant jurisdiction or does not respond, and removed content may later reappear elsewhere.

American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (2014). Standards for educational and psychological testing. Washington DC: AERA.

American Psychological Association. (2017). Ethical principles of psychologists and code of conduct (2002, amended effective June 1, 2010, and January 1, 2017). https://www.apa.org/ethics/code/

International Test Commission. (2013). The ITC guidelines on test use. International Test Commission. Retrieved from https://www.intestcom.org/files/guideline_test_use.pdfopens in new window

International Test Commission. (2014). The ITC guidelines on the security of tests, examinations, and other assessments. International Test Commission. Retrieved from: https://www.intestcom.org/files/guideline_test_security.pdfopens in new window

Tippins, N. T. (2015). Technology and assessment in selection. Annual Review in Organizational Psychology and Organizational Behavior, 2, 551-582. doi: 10.1146/annurev-orgpsych031413-091317