Encryption: the burden of proof

The economic cost of a backdoor · Part 7 of 7

A mandate for access to encrypted messages should be judged by the useful information it delivers, the risks it adds and the resources needed to oversee it. From the FBI’s data collection to China’s compliance requirements and Europeans’ views on privacy, this final instalment examines the evidence a public decision needs.

On 21 October 2026, the US Federal Bureau of Investigation (FBI) published an administrative notice seeking renewal of its Lawful Access Data Collection. The programme gathers reports of investigations affected by encryption. Comments remain open until 20 November. A month earlier, the agency had acknowledged limited participation and little reported data. Establishing the scale of the investigative problem remains a work in progress. ^X01

The notice concerns voluntary statistical reporting. It offers a concrete starting point for the question that closes this series: what evidence should justify requiring access to encrypted data? An investigative obstacle, a technical solution and a public benefit each need examination. Connecting them takes work, explicit assumptions and institutions capable of checking the results. ^{X01 X04}

Here, the burden of proof means that requirement for public evaluation. An access mandate should be specific enough to describe, compare with available alternatives and reassess in operation. This is the final instalment’s editorial proposition, informed by American, European and Chinese documents. It applies to the benefits sought as well as the risks others would bear.

From reported obstacles to investigative outcomes

The FBI began accepting LADC submissions on 1 January 2024. Its launch bulletin reported the absence of a nationally representative count. The original methodology described voluntary, incident-based submissions containing information about devices and access attempts. The 2026 notices confirm voluntary participation and, crucially, warn about its limited extent. ^{X02 X01}

That limitation changes how a figure should be read. Agencies best equipped to report may also handle the most complex cases. One case can involve several devices. A technical obstacle might delay an investigation, leave a question unanswered or be overcome through other evidence. Diagnosis requires the counting unit, the collection’s coverage and what subsequently happened to the case. The US Congress’s audit institution, the Government Accountability Office (GAO), frames data reliability in its guide through accuracy, completeness and suitability for the intended use. We apply that method here; the GAO document does not audit LADC. ^{X04 X06}

An American precedent explains the concern. In a note appended to a 2017 encryption speech, the Justice Department acknowledges that a figure of 7,800 devices had been cited incorrectly because of an error in FBI methodology. The corrected number would be substantially lower, it says, without supplying it. The page does not establish when the correction was inserted. A number can acquire considerable influence before its construction is settled. ^X03

The decisive measure would be the additional benefit attributable to a specified policy. Consider a category of investigations defined in advance: comparable offences and deadlines, documented resources. Would a new capability deliver useful information sooner? Would it change a decision, corroborate a lead or clear someone of suspicion? How much would existing means have achieved? These questions extend the National Academies’ evaluation framework, which separates inaccessible devices from their implications for investigations. ^X04

The comparison will be imperfect. Cases differ, agencies learn and targets may change their behaviour. A credible evaluation must explain how it handles those differences. It could use anonymized retrospective case reviews, controlled technical tests and independent scrutiny of assumptions. The appropriate method depends on the proposed system and legal constraints.

The strongest case for access

The content of a conversation can change the course of an investigation. In a contribution published on 4 May 2026, France’s Parliamentary Delegation for Intelligence argues that message content can be essential to establish the role of an organizer or a more distant accomplice. Connections between correspondents do not always reveal what their instructions mean. The delegation says targeted computer-data collection requires time, expertise and resources that constrain its use, particularly in emergencies. ^X11

This is a substantive argument. Information that arrives after an operational deadline may lose much of its value. An assessment confined to computing costs would miss the harm associated with a delayed investigation or a poorly understood threat. The National Academies’ framework explicitly considers whether access is available soon enough and at the required scale. ^X04

Part of the delegation’s underlying work is classified. Its public contribution supplies no sample from which to estimate how many additional cases a particular architecture would resolve. It identifies an operational need and advocates targeted, supervised access. Turning that need into policy still requires a description of the capability, its intended users and its scope. ^X11

Precision also improves the debate. Obtaining data already held by a provider, accessing a device and changing a service to make content accessible involve different technical arrangements. The word “access” leaves those differences unresolved. The National Academies recommend assessing each proposal through its capabilities, security implications, effects on rights, costs and international context. ^X04

Testing the whole system

Assessing a proposal means following authorization as closely as cryptography. Who can initiate access? Which devices and software versions are covered? How would the system detect an erroneous order, a compromised administrator account or use beyond the authorized scope? What oversight remains as the service changes? Integration and operation need examining alongside the promised capability. ^{X04 X08}

The cryptographic-module validation FAQ from the US National Institute of Standards and Technology (NIST) provides a precise reference point. A module is evaluated within a defined validation boundary. Embedding it in a product does not automatically validate the product or its use of the module. Validation remains valuable within that scope. Our inference is that an access mechanism also requires evaluation of the surrounding system: authorization decisions, audit records and maintenance procedures. ^X05

The European Data Protection Board made a related demand in November 2024: specify and assess concrete technical measures to determine their necessity, proportionality and effectiveness. It observes that a legally targeted intervention can require a technical change with broader reach. It calls for objective evidence, including relevant statistics, while recognizing the concern that certain offences may go unpunished. ^X08

The European timetable shows why this work needs to be documented. A June 2025 Commission communication envisaged a technology roadmap on encryption. A January 2026 Council note expected results in the fourth quarter of 2026, following an initial second-quarter target. By 7 November, we had found no public confirmation of final publication. A policy deadline describes the work programme; the technical case depends on the results available for examination. ^X09

The construction of the $250.6 billion AES estimate

The economics of encryption offers an instructive example. A study commissioned by NIST and published in September 2018 assessed the programme behind AES, an encryption standard. It estimated $250.6 billion in gross benefits, accumulated from 1998 to 2017 in constant 2017 dollars. The estimate concerns the AES programme over 1996–2017. ^X19

Understanding the result requires examining its comparison scenario. The authors estimated development, adoption and interoperability costs that organizations would have incurred without NIST’s programme. Their counterfactual describes the industry’s path towards an alternative standard. This alternative history permits an estimate of avoided costs, while requiring assumptions about companies’ choices and timing. ^X19

The survey supplied information for direct estimates from 74 respondents. A model supplied estimates for another 95, bringing the initial group to 169. The authors then extrapolated to selected sectors using respondents’ share of employment, retaining sectors with at least ten respondents. Past benefits were accumulated to 2017 at a 7% real annual rate. Each stage adds an assumption to the calculation. ^X19

Method and scope

Leech, Ferris and Scott, NIST GCR 18-017, September 2018. Program evaluated: 1996–2017; underlying benefits: 1998–2017. Survey from 26 February to 9 April 2018: 74 respondents supplied inputs for direct estimates, with 95 more estimated through a model, totalling 169. Sector extrapolation uses the employment share represented, retaining sectors with at least ten respondents. Estimated gross benefits are $250.6 billion in constant 2017 dollars, accumulated to 2017 at a 7% annual real rate. Arrows describe calculation stages, without proportional widths or areas. Appendix E recognizes unquantifiable initial uncertainty; results could be much lower or higher. The estimate covers the AES program, rather than all encryption or a loss attributable to a backdoor.

The uncertainty appendix matters as much as the headline figure. The authors acknowledge that uncertainty in the initial estimates, which they consider the most important, cannot be quantified. Benefits could be much lower or higher. Describing the result as a “minimum” reflects excluded benefit categories and sectors; it does not establish a guaranteed statistical lower bound. ^X19

The methodological lesson is useful: shared infrastructure can save organizations from rebuilding the same solutions, but estimating that benefit requires a plausible replacement scenario. The AES total remains tied to this scenario, period and survey. Applying an arbitrary percentage to it to price a backdoor would produce an unsupported number.

A specified access mandate would require the same effort of definition: the changes imposed, resources consumed and events capable of causing harm. Design costs can be documented through quotations and working time. Expected losses also require assumptions about incident probabilities and severity. When those parameters are unknown, an explicitly hypothetical sensitivity analysis can show what the result depends on. Presenting the resulting total as observed would mislead.

The value people place on confidentiality

A survey by the European Union Agency for Fundamental Rights (FRA), published on 11 May 2026, supplies a different kind of evidence. Between 6 and 30 November 2025, 26,453 people aged 15 and over across the European Union answered questions about private online communications. Across all respondents, including non-users, 77% considered encryption very or fairly important. The results are weighted. ^X07

The concern question specifies different actors and access without the person’s knowledge or permission. In France, 64% expressed concern about people engaged in criminal activity, 50% about companies and 33% about law enforcement. The figure presents all six categories on a consistent basis alongside the EU aggregate, which includes France. ^X07

Method and scope

FRA publication of 11 May 2026, Eurobarometer 104.2; fieldwork from 6 to 30 November 2025. Weighted percentages of people aged 15 and over, including non-users, giving concern scores of 5–7 on a 1–7 scale. The EU27 sample contains 26,453 respondents; that total does not describe the French sample. France then EU: people involved in criminal activity 64/62%, foreign governments 49/49%, businesses 50/47%, national intelligence 37/42%, national government 41/42%, law enforcement 33/37%. The common axis runs from 0 to 100%. France is included in the EU aggregate; responses overlap and cannot be added. Comparisons are descriptive, with no supplied confidence intervals for these gaps. The question measures neither support for a specific judicial obligation nor willingness to pay.

These answers describe expectations and concerns. They establish neither willingness to pay nor a view on a defined judicial-access mandate. Someone can value confidentiality highly and approve of certain forms of access under conditions. Differences of a few percentage points between France and the EU remain descriptive: the report reviewed supplies no confidence intervals for these comparisons. ^X07

An economic claim therefore needs another step. To establish that a particular architecture would cost a provider customers, an analysis would have to observe or estimate their response to that architecture against a credible comparison. The survey supplies no such elasticity. It helps identify the safeguards people may expect, a question that must then be connected to purchasing decisions, usage and rights.

The distribution of consequences creates another difficulty. In 2006, Ross Anderson and Tyler Moore explained how security incentives become distorted when the actor choosing protection differs from the actor bearing the loss. Applied here, their framework suggests tracing separately the authority demanding a capability, the company building it, the agencies using it and the people whose data could be exposed. This is an analytical application, without an estimate of current losses. ^X20

Public reimbursement can shift a bill between a provider and taxpayers. The working hours are still consumed. Compensation for victims, if available, follows the harm. Distinguishing financial transfers from resources used helps separate the government budget, company spending and costs to society. It also raises the question of who would fund continuing risk reduction. ^{X20 X04}

Safeguards need staff

France’s National Commission for the Control of Intelligence Techniques (CNCTR) gives that question a practical form in its 2025 annual report. It describes more technical work with unchanged resources and considers itself at its capacity limit. Its scrutiny follows data through collection, use, traceability and destruction. This assessment belongs to intelligence oversight and its particular legal framework; it illustrates the human work behind a safeguard. ^X10

The CNCTR reports 120 inspections and 100,813 requests for individual domestic intelligence techniques in 2025. Dividing one by the other would create a meaningless coverage rate: an inspection can examine several techniques, while requests include renewals. The economic question lies in the depth of scrutiny, the access granted to inspectors and the expertise required. The commission also observes that closer examination can uncover more anomalies. Any rise in detected problems must be interpreted alongside the intensity of oversight. ^X10

The FRA’s October 2026 report on data protection in policing and justice adds another perspective. Interviewees describe needs for training, expertise and resources among those responsible for safeguards. This qualitative research comprises 132 interviews conducted between January and July 2025. It supplies no representative non-compliance rate. No operational police or judicial interview was secured in France; the European findings retain that limitation. ^X13

China supplies another documentary example, in a different field: compliance in the use of cryptography. Its 2023 regulations define commercial cryptography through the protection and authentication of information outside the state-secrets category. They call for coordination between assessments to avoid duplication. That requirement establishes an organizational objective; savings achieved would still need to be measured. ^X15

New rules for the relevant electronic authentication providers took effect on 1 July 2026. They require compliance assessment at least annually, remediation and submission of the report to the competent authority. Relevant technical and security-management staff must receive at least 20 hours of training each year. Administrative inspections are free of charge. Both provisions can apply together: an inspection without a fee leaves the operator with duties that consume working time. ^X14

Other texts describe the oversight process. For certain important networks and systems subject to commercial-cryptography requirements, assessment begins at design, continues before operation and recurs at least annually. Self-assessment is permitted under conditions. Testing bodies face independence and record-retention requirements; failure in certain checks requires remediation and suspension of the affected testing activity. ^{X16 X17}

A Beijing notice dated 21 January 2026 adds a detail that matters for information quality: results must be filed even when the assessment fails, within thirty days of the report’s completion. This gives the administration a channel for receiving adverse results. Filing alone establishes nothing about public disclosure, completeness of reporting or effective correction of defects. ^X18

These Chinese documents describe security duties within defined fields. They establish no general ability to decrypt communications. Their comparative value lies in the recurring work of oversight: conduct an assessment, preserve records, address failures and repeat the process. Safeguards, rights and institutional independence remain specific to each regime. ^{X14 X16 X17 X18}

Build in a chance to reconsider

A European precedent shows how monitoring can be written into legislation. The electronic-evidence regulation requires statistics on orders, execution, refusals, timing, remedies, and costs claimed or reimbursed. Evaluation is due by 18 August 2029. The regulation concerns production of data and expressly creates no duty to decrypt. Its monitoring arrangements are useful here as a precedent from a field distinct from exceptional access. ^X12

Such statistics can describe activity and some costs. Assessing another policy’s additional benefit would require evidence of its actual usefulness and the alternatives available. Evaluation would also need to consider incidents, disputed uses and consequences for people outside the investigations. Continuing scrutiny is part of the National Academies’ framework. ^X04

Drawing on these documents, l0g proposes a decision rule: match the breadth of an obligation to the strength of the evidence, then arrange for reassessment. That begins with a technical description precise enough for independent examination. It requires a documented need, comparison with other means and analysis of who bears the risk.

Funding safeguards belongs in the initial decision. Appointing an oversight body gives it a task; access, expertise and resources enable it to carry that task out. Procedures for correcting, suspending or withdrawing a system should also be prepared before deployment. A change in the product or the threat can alter the original balance. This is our recommendation, informed by the monitoring and evaluation requirements in the sources. ^{X04 X08 X10}

Method and scope

Decision framework proposed by l0g, drawing on chapter 7 of the 2018 National Academies report, EDPB Statement 5/2024 and monitoring provisions in Regulation (EU) 2023/1543. Branches show options to prepare; they depict neither a universal legal procedure already in force nor a validated technical architecture. The European regulation supplies a separate example of statistics and evaluation for data production, with an evaluation deadline of 18 August 2029; it creates no decryption obligation. The diagram contains no measured frequency, probability or duration. Technical scope, safeguards and deployment conditions require their own assessment.

The public record examined here documents investigative needs, risk mechanisms, duties and some resources committed. It leaves open the causal benefit of an architecture yet to be specified and its full social cost. A useful number first requires a defined proposal, comparison scenario, time horizon and affected population.

Advocates of a new access capability must demonstrate its contribution and safeguards. Those proposing to forgo it must examine the cost of lost information and the actual reach of alternatives. Serious public decision-making makes both arguments available for scrutiny, with their assumptions and limits. It then equips itself to recognize a fulfilled promise, an unnecessary expense or a risk that has become too great.

---

Method and documentary cut-off. Research on public sources, current to 7 November 2026, Paris time. The twenty references distinguish publication dates, events, periods and documentary scope. Statements by administrations, NIST and parliamentarians are attributed; Chinese requirements remain distinct from observed practice. The original LADC methodology comes from an FBI document archived by the unofficial OMB.report mirror. The form actually used in November 2026 remains outside the verified material. The decision framework is this analysis’s proposal.

The six preceding parts

The French parliamentary debate opens the series. The parts on devices and backups and oversight of state access follow data and authorizations.

The analysis of iCloud and borders examines safeguards and switching costs. Damage after hijacked access and alternative routes to evidence complete the comparison of resources and risks.

The sharing image balances a protected message against material submitted for examination. This conceptual composition illustrates the demand for evidence; it measures no economic balance and depicts no actual court procedure.

Sources

1. X01 · Federal Bureau of Investigation, Department of Justice / Federal Register, Government Publishing Office · 2026-10-21 · Accessed 7 November 2026Lawful Access Data Collection : renewal notice of 21 October 2026 and preceding notice of 18 September
   Passages and scope

   Federal Register, FBI notices dated 18 September and 21 October 2026, OMB 1110-0088, DATES, Abstract and Overview. LADC reporting is voluntary; the September notice acknowledges limited participation and reported data. The later notice seeks PRA renewal and accepts comments until 20 November. Administrative respondent and burden estimates are distinct from observed investigations and policy outcomes. The later notice refers to 19 September for the previous publication, whose official issue is dated the 18th.

   Additional document

   Additional document

2. X02 · Federal Bureau of Investigation, CJIS/UCR; original FBI document archived by the unofficial OMB.report mirror · Accessed 7 November 2026Lawful Access Data Collection : January 2024 launch and original collection methodology
   Passages and scope

   FBI UCR Program Quarterly 24-1, January 2024, printed pages 3 to 5: reporting opened on 1 January, with no nationally representative count then available. The initial July 2023 Supporting Statement A describes device and access-attempt variables and voluntary incident-by-incident reporting. The FBI document was read through a copy hosted by OMB.report, an unofficial archive; its Reginfo route remains inaccessible. The original design establishes neither the precise form used in November 2026 nor representative coverage of submitted reports.

   Additional document

   Additional document

   Additional document

   Additional document

3. X03 · US Department of Justice, Office of Public Affairs · 2017-10-10 · Accessed 7 November 2026Rod J. Rosenstein’s remarks on encryption at the United States Naval Academy : official methodological correction
   Passages and scope

   DOJ speech dated 10 October 2017, final double-asterisk note. The department corrects an earlier count of 7,800 devices, attributing it to an error in FBI methodology, and says the correct total will be substantially lower without giving it. The displayed 5 February 2025 web update does not necessarily date the correction. This establishes a historical correction, without a final replacement figure, overstatement rate or current measure of investigative obstacles.

4. X04 · National Academies of Sciences, Engineering, and Medicine / The National Academies Press · Accessed 7 November 2026Decrypting the Encryption Debate: A Framework for Decision Makers : Chapter 7, A Framework for Evaluating Approaches to Access Plaintext
   Passages and scope

   National Academies consensus report, 2018, chapter 7, printed pages 83 to 92, especially Box 7.2 and the eight evaluation questions. The framework examines capability, timeliness, scale, security, rights, costs, commerce and continuing oversight. It calls for assessing component integration and consequences of both authorized use and misuse. Behavioral projections and incremental effects are difficult to quantify. The historical framework analyzes trade-offs; it neither decides whether exceptional access should be required nor supplies its overall cost.

   Additional document

   Additional document

5. X05 · National Institute of Standards and Technology, Computer Security Resource Center · Accessed 7 November 2026Cryptographic Module Validation Program : NIST FAQ on the scope of FIPS 140 validation
   Passages and scope

   NIST Cryptographic Module Validation Program FAQ, P-16 to P-18. The certificate and security policy define a module’s validation scope; a product incorporating that module does not itself become validated. Correct use of the embedded module falls outside that assurance. These statements concern FIPS 140 and a defined technical boundary. Applying this distinction to exceptional-access authorization, integration and operations is the article’s analysis, separately supported by the National Academies framework.

6. X06 · US Government Accountability Office · 2019-12-16 · Accessed 7 November 2026Assessing Data Reliability : GAO-20-283G
   Passages and scope

   GAO-20-283G, published 16 December 2019, official Fast Facts and Highlights. The guide evaluates data reliability through accuracy, completeness and applicability to an audit’s purpose, using a risk-based framework adapted to the engagement. It is general guidance, rather than a GAO audit of LADC or validation of the FBI figures. Its application to encryption-related reporting is the article’s methodological choice.

   Additional document

7. X07 · European Union Agency for Fundamental Rights (FRA) · 2026-05-11 · Accessed 7 November 2026Encryption and concerns about access to private online communication : people’s views across the EU
   Passages and scope

   FRA, published 11 May 2026, methodology pages 4–5 and Table 1 pages 10–11. Fieldwork: 6–30 November 2025; 26,453 respondents aged 15+, weighted results including non-users. Encryption importance: 77%. Concern, France/EU27: criminals 64/62, foreign governments 49/49, businesses 50/47, intelligence 37/42, national government 41/42, police 33/37%. Answers overlap; France belongs to the EU aggregate. Reported views establish neither willingness to pay nor support for a specified mandate.

   Additional document

   Additional document

8. X08 · European Data Protection Board (EDPB) · 2024-11-04 · Accessed 7 November 2026Statement 5/2024 on the Recommendations of the High-Level Group on Access to Data for Effective Law Enforcement
   Passages and scope

   EDPB Statement 5/2024, adopted 4 November, pages 2 and 4–6. The board calls for objective evidence and defined technical measures to assess necessity, proportionality and feasibility. It distinguishes legal targeting from the technical reach of changes and examines whole communications and endpoints. This institutional assessment addresses the recommendations reviewed; it neither experimentally validates nor disproves every conceivable architecture.

9. X09 · European Commission; Council of the European Union follow-up note · 2025-06-24 · Accessed 7 November 2026Roadmap for effective and lawful access to data for law enforcement : COM(2025) 349 final and Council follow-up 5424/26
   Passages and scope

   Commission communication COM(2025) 349, 24 June 2025, sections IV–V; Council note 5424/26, 23 January 2026, page 5, D1. The initial roadmap targeted the second quarter of 2026; the follow-up expected results in the fourth quarter. The official Encryption page still presents prospective work. Official-source searches through 7 November identified no confirmation of final publication. Plans and deadlines remain distinct from a validated, deployed technical solution.

   Additional document

   Additional document

10. X10 · France’s National Commission for the Control of Intelligence Techniques (CNCTR) · 2026-06-25 · Accessed 7 November 2026CNCTR Annual Activity Report 2025
    Passages and scope

    CNCTR’s 2025 report, published 25 June 2026, pages 21–25, 38 and 58–60: 100,813 individual domestic intelligence-technique requests and 120 inspections; the commission reports a capacity limit with unchanged resources. Requests include renewals. The 2025 budget performance report, indicator 5.3, page 148, counts one visit even when several techniques are inspected and adds some remote checks in 2026. These units preclude a 120/100813 coverage rate. This is the institution’s assessment within intelligence oversight.

    Additional document

    Additional document

11. X11 · Parliamentary Delegation for Intelligence / French National Assembly and Senate · 2026-05-04 · Accessed 7 November 2026French Parliamentary Delegation for Intelligence : communication on encryption and algorithms
    Passages and scope

    French Parliamentary Delegation for Intelligence, published 4 May 2026, pages 2–4. The delegation argues that message content can establish particular criminal roles and describes the staffing, complexity and urgency constraints of targeted device access. Its case for supervised access draws partly on classified work. The public contribution documents an operational position; it supplies neither a demonstrated architecture nor a sample allowing additional benefits to be calculated.

    Additional document

12. X12 · European Parliament and Council / Official Journal · 2023-07-28 · Accessed 7 November 2026Regulation (EU) 2023/1543 : costs, monitoring and evaluation of European electronic-evidence orders
    Passages and scope

    Regulation (EU) 2023/1543, 12 July 2023, Articles 14, 28 and 33–34, recital 20. Statistical monitoring from 18 August 2026 covers execution, refusal, timing, remedies and costs claimed or reimbursed; evaluation is due by 18 August 2029. Prior-year data must be transmitted before 31 March. The regulation concerns production of data and creates no decryption duty. Statutory deadlines and requirements remain separate from observed implementation.

    Additional document

13. X13 · European Union Agency for Fundamental Rights (FRA) · 2026-10-09 · Accessed 7 November 2026The Law Enforcement Directive : data protection challenges in practice
    Passages and scope

    FRA, published 9 October 2026, findings and Annex 1, pages 81–84. Qualitative research: 132 interviews, detailed fieldwork January–July 2025, comprising 42 data-protection officers, 61 operational staff and 29 supervisory-authority staff. The annex records no operational police or judicial interview in France. The reported training and resource needs establish neither a representative non-compliance rate nor a measurement specifically concerning backdoors.

    Additional document

14. X14 · State Cryptography Administration, China · 2026-06-03 · Accessed 7 November 2026Rules on the Use of Cryptography by Electronic Authentication Service Providers : State Cryptography Administration Order No.6
    Passages and scope

    SCA Order 6, published 3 June 2026, effective 1 July; Articles 2–3, 16–17 and 20. Licensed electronic-authentication providers must assess compliance at least annually, correct problems and report to their provincial authority. Relevant technical and security-management staff require at least 20 training hours annually. Administrative inspections carry no fee. These duties have a defined scope and supply no measurement of compliance, total costs or a general decryption capability.

    Additional document

15. X15 · State Council of China; official reproduction by the Ministry of Ecology and Environment · 2023-05-29 · Accessed 7 November 2026Regulations on Commercial Cryptography : State Council Decree No.760
    Passages and scope

    China’s Commercial Cryptography Regulation, Decree 760, promulgated 27 April 2023, effective 1 July; official MEE reproduction dated 29 May. Articles 2, 42 and 47 cover non-state-secret information, coordination to avoid duplicate assessments and protection of proprietary cryptographic information. Prescribed coordination differs from measured savings. The legal scope neither encompasses all private communications nor establishes general access to their content.

16. X16 · State Cryptography Administration, China · 2023-10-07 · Accessed 7 November 2026Measures for Security Assessments of Commercial Cryptography Applications : State Cryptography Administration Order No.3
    Passages and scope

    SCA Order 3, published 7 October 2023, effective 1 November; Articles 6–9 and 12–14. Certain important networks and systems require assessment during design, before operation and at least annually thereafter. Conditional self-assessment, filing within thirty days and formal review are provided. Failure before operation requires remediation before launch; during operation, remediation and security safeguards apply. These duties supply no measurement of actual compliance or observed cost.

17. X17 · State Cryptography Administration, China · 2023-10-07 · Accessed 7 November 2026Measures for Commercial Cryptography Testing Bodies : State Cryptography Administration Order No.2
    Passages and scope

    SCA Order 2, published 7 October 2023, effective 1 November; Articles 15–19 and 23. Testing bodies must maintain independence and traceability, retain original records and reports for at least six years, and address failed prescribed checks. Remediation lasts at least six months, suspending the affected testing scope until subsequent approval. That suspension concerns testing activity, separate from operation of assessed networks. No implementation rate is established.

18. X18 · Beijing Cryptography Administration · 2026-01-21 · Accessed 7 November 2026Beijing notice on filing the results of commercial cryptography application security assessments
    Passages and scope

    Beijing Cryptography Administration notice, 21 January 2026, sections I–II. Covered operators must file assessment results, whether successful or failed, within thirty days of report completion. Formal review and sample checks apply; a failed formal submission must be filed again. This local administrative procedure establishes neither public disclosure nor complete reporting or effective remediation.

19. X19 · David P. Leech, Stacey Ferris and John T. Scott; NIST-commissioned study · 2018-09-07 · Accessed 7 November 2026The Economic Impacts of the Advanced Encryption Standard, 1996–2017 : NIST GCR 18-017
    Passages and scope

    Leech, Ferris and Scott, contracted NIST GCR 18-017 study, September 2018, sections 5.3.3 and 6 and Appendix E. Estimates derived from 74 responses are supplemented by modeled benefits for 95 others, totaling 169. Employment-share extrapolation includes sectors with at least ten respondents. US$250.6 billion is estimated gross benefit in constant 2017 dollars, accumulated from 1998 at a 7% real rate. The program evaluated spans 1996–2017. Principal uncertainty remains unquantifiable; the minimum framing is not a guaranteed statistical bound. The figure estimates no backdoor-related loss.

    Additional document

    Additional document

20. X20 · Ross Anderson and Tyler Moore / Science, vol. 314, no. 5799, pp. 610–613 · 2006-10-27 · Accessed 7 November 2026The Economics of Information Security
    Passages and scope

    Anderson and Moore, Science 314(5799), 27 October 2006, pages 610 to 613; fourteen-page author copy, introduction and security-externalities section. The paper examines incentives when security decision makers differ from those bearing failure costs, and interdependence between investments. This historical economic framework supports analysis of how consequences are distributed. It supplies no current price, probability or loss for a specified access architecture.

    Additional document