Whose Knowledge Is It? The Legality, Consequences & Governance of Employer and Government Capture of Worker Expertise via AI. The new practice of recording how experts work & using it to train AI...
...that could replace them sits in a contested legal grey zone touching data-protection law, the EU AI Act, and human-rights instruments—none of which clearly prohibits it. The threat is real.
Summary: The report finds that capturing a worker’s general skills has always been legal, but the new practice of recording how experts work and using it to train AI that could replace them sits in a contested legal grey zone touching data-protection law, the EU AI Act, and human-rights instruments—none of which clearly prohibits it.
Economically, the threat is real: AI trained on top performers compresses the wage premium on experience and, under current ownership rules, shifts bargaining power from workers to firms, making both feared scenarios (selling captured expertise; not hiring or training juniors) plausible extensions of documented trends.
The most effective remedy is collective rather than individual—individual data ownership can backfire, whereas collective bargaining over worker data, revenue-sharing, co-determination, and WGA/SAG-AFTRA-style AI clauses offer workable templates for unions and regulators.
Whose Knowledge Is It?
The Legality, Consequences, and Governance of Employer and Government Capture of Worker Expertise via AI
Research briefing
by Claude
22 July 2026
Executive summary
The premise that a worker’s expertise can now be “captured” by employers and governments through AI platforms is only half right. Capturing an employee’s general skills, knowledge, and experience has always been lawful: courts across the United States, the United Kingdom, and the European Union decline to treat general expertise as protectable intellectual property, and the “work made for hire” doctrine already vests employers with copyright in most work product. What is genuinely new and legally contested is the systematic recording of how expert workers do their jobs and the repurposing of that behavioural record to train AI systems that can replicate — and potentially replace — them. This practice runs up against data-protection law (purpose limitation and automated-decision rules under the GDPR), the right to private life (Article 8 of the European Convention on Human Rights), the EU AI Act, and international human-rights instruments. None of these clearly prohibits it, but several constrain it.
The labour-market threat is real and empirically grounded. Peer-reviewed and working-paper evidence shows that AI trained on the behaviour of top performers disseminates their tacit knowledge to novices, compresses the wage premium that experience commands, and — under current ownership rules — shifts bargaining power from workers to firms. Firms can monetize a worker’s captured expertise, and can decline to hire or train juniors because a senior’s expertise has been copied into an AI agent.
The most effective remedy is collective rather than individual. The leading economic model finds that giving workers individual ownership of their data can backfire, because workers’ data are substitutes: each individual sale weakens the position of everyone else. Collective data ownership or bargaining, by contrast, can restore both efficiency and worker welfare. Concrete templates already exist, including the Writers Guild and SAG-AFTRA AI clauses, the German co-determination model, the EU Platform Work Directive, and the Trades Union Congress’s model AI Bill.
Key findings
● Skills are not property; work product often is. The universal common-law rule is that an employee’s general knowledge, skill, and experience cannot be a trade secret and can be carried to a new employer. But copyright’s work-made-for-hire doctrine vests authorship of in-scope work product in the employer, and assignment clauses capture inventions. The AI-capture concern sits in the gap between these: the behavioural record of expert work is neither classic intellectual property nor classic “skill.”
● Data-protection law is the sharpest existing constraint — but it was not built for this. Under the GDPR and UK GDPR, employee consent is generally not “freely given,” so employers rely on “legitimate interests,” which must pass a balancing test and respect purpose limitation and data minimisation. Repurposing monitoring data collected for “quality” or “security” into AI training data is a purpose-limitation problem. Article 22 restricts solely-automated decisions with significant effects.
● Human-rights law recognises a right to the fruits of authorship — but excludes mere skills and rarely binds private employers. The Universal Declaration and the ICESCR protect the interests of authors of scientific, literary, or artistic production; the relevant treaty commentary ties this to human creators and an adequate standard of living. These protect creative and scientific output, not tacit job know-how, and mostly bind states rather than private firms.
● The EU is regulating; the US is largely not. The EU AI Act makes most employment AI “high-risk” (worker notice, human oversight, logging) and bans workplace emotion-recognition. The Platform Work Directive regulates algorithmic management. The US relies on narrow state and city rules and self-regulation; the federal non-compete ban was struck down.
● AI compresses the skill premium. A study of 5,172 customer-support agents found that AI assistance raised productivity (issues resolved per hour) by about 15% on average — roughly a 34% gain for novice and low-skilled workers but minimal impact for the most experienced. The authors describe the tool as one that “disseminates the best practices of more able workers.” This is the empirical core of the concern: the senior worker’s tacit edge is transferred to everyone.
● Collective ownership beats individual ownership. A formal economic model finds that individual data ownership eliminates knowledge-withholding but creates a negative externality — “one worker’s data strengthens the firm’s bargaining position against others” — while collective data ownership can achieve the efficient, welfare-improving outcome.
Part (A) — Is capturing a worker’s expertise legal and ethical?
Intellectual property: the “skills gap”
The foundational principle, recognised in every US state and echoed across the UK and EU, is that an employee’s general knowledge, skill, and experience cannot be protected as a trade secret. Former employees are free to use the general knowledge, skill, and experience acquired on the job without incurring liability for misappropriation — a rule tied directly to protecting competition and employee mobility. Firms have always benefited from, and workers have always been free to carry away, on-the-job expertise.
The offsetting doctrine is copyright’s work made for hire. Under the US Copyright Act, the employer is treated as the author of work created within the scope of employment; combined with invention-assignment agreements and trade-secret protection for genuinely secret information, employers already own most tangible work product. The novel question is the behavioural training data — call transcripts, keystrokes, screen recordings — which is neither the employee’s transferable “skill” nor a discrete copyrightable “work.” This is the legal grey zone that can be identified.
Data protection and privacy
Consent is weak. European data-protection authorities hold that, because of the dependency in the employment relationship, an employee can rarely give consent that is genuinely free. Employers therefore rely on “legitimate interests,” subject to a documented balancing test, or contractual necessity.
Purpose limitation and minimisation. Data lawfully collected for monitoring, quality, or security cannot automatically be repurposed for AI training; the GDPR requires a fresh compatibility assessment, and systematic monitoring triggers a Data Protection Impact Assessment. Covert monitoring is heavily restricted.
Automated decisions. Solely-automated decisions producing legal or similarly significant effects — on performance, discipline, or termination — are restricted absent narrow exceptions and safeguards including meaningful human review; European case law confirms that a rubber-stamp review does not count as human involvement. Several member states (Germany, France, Italy) impose stricter workplace rules. The United States has no federal baseline; state laws such as California’s give some access and deletion rights but were not designed for workplace AI training.
Human rights
The Universal Declaration of Human Rights (Article 27) and the International Covenant on Economic, Social and Cultural Rights (Article 15) protect the moral and material interests resulting from a person’s scientific, literary, or artistic production. The authoritative treaty commentary stresses that this is a human right of human authors — it derives, in its words, from “the inherent dignity and worth of all persons” — that corporations are excluded as “authors,” and that authors’ material interests are linked to an adequate standard of living. Crucially, the right protects creative and scientific productions, not tacit job know-how.
The European Convention on Human Rights protects private life (Article 8). In Bărbulescu v Romania (2017), the Grand Chamber held that Romania violated Article 8 by failing to protect an employee dismissed after his messages were monitored, reasoning that “an employer’s instructions cannot reduce private social life in the workplace to zero” and that courts must weigh prior notice, the scope of monitoring, less-intrusive alternatives, and the specific justification. The EU Charter of Fundamental Rights adds rights to private life and communications (Article 7), to protection of personal data processed fairly for specified purposes (Article 8), and to working conditions that respect dignity (Article 31). At the international level, the International Labour Organization’s 2019 Centenary Declaration calls for privacy and data protection and a “human-centred approach” to the digital transformation of work; the ILO’s 2019 Global Commission separately urged a “human-in-command” approach to technology. A 2025 ILO report addresses algorithmic management directly but is analytical rather than a binding standard.
Constitutional dimensions
US constitutional protections mostly bind government, not private employers. For public employees, the Supreme Court held in O’Connor v Ortega (1987) that “individuals do not lose Fourth Amendment rights merely because they work for the government,” but applied a lenient reasonableness test that balances the employee’s privacy expectation against the government’s operational needs, rather than requiring probable cause. Citizens interacting with government AI systems can invoke constitutional and administrative-law protections (due process, and in the EU, automated-decision rules and the AI Act’s high-risk requirements for public services). A “takings” theory treating expertise as property is largely untested and doctrinally weak, precisely because skills are not recognised as property.
The EU AI Act and emerging regulation
Employment AI — for recruitment, evaluation, promotion, termination, or task allocation — is classified as high-risk under the EU AI Act (Regulation (EU) 2024/1689). From 2 August 2026, deployers must ensure human oversight and logging and must inform workers’ representatives and affected workers before such a system is used. Emotion-recognition in the workplace is prohibited from 2 February 2025, except for medical or safety reasons. Fines can reach €15 million or 3% of turnover. (A November 2025 “Digital Omnibus” proposal may adjust some deadlines, but it is a proposal, not enacted law.) The Platform Work Directive (Directive (EU) 2024/2831, in force since December 2024, with transposition due by December 2026) mandates algorithmic transparency, requires human review of significant decisions, bans purely automated dismissal, and prohibits processing data on the emotional or psychological state of platform workers or data used to predict the exercise of fundamental rights, including collective bargaining.
US state and city laws
New York City’s Local Law 144 (enacted 2021; enforced from July 2023) requires annual independent bias audits of automated employment decision tools, at least ten business days’ candidate notice, and public posting of audit summaries. Penalties run from $500 for a first violation to $500–$1,500 for each subsequent violation, with each day of use counted separately. Illinois’ AI Video Interview Act requires notice and consent; Colorado and other states are following. These measures target bias, not expertise capture.
Ethics
Commentators frame expertise capture as a dignity and exploitation problem. Labour-law analysts argue that capturing tacit expertise through surveillance and repurposing it for automation treats workers as instruments for extracting their own know-how, and that mandatory process transparency can convert a worker’s craft knowledge into a surveillance substrate that accelerates deskilling and erodes bargaining power. The historical analogy is Taylorist scientific management, which sought to surveil skilled workers and codify their techniques — and which provoked sustained labour resistance a century ago.
A counterargument worth stating
Sceptics argue that this is not fundamentally new. Firms have always learned from their workers, apprentices have always “trained their replacements,” and the real variable is labour’s bargaining power rather than AI as such. In this view, the most effective “AI policy” for workers is to strengthen their bargaining position — through unionisation, competition policy, and limits on restrictive covenants — rather than to micromanage the technology.
Part (B) — Mobility and the consequences for employees
The core mechanism
The clearest formalisation of the concern comes from the working paper Labor as Capital: AI and the Ownership of Expertise (Cullen, Li and Li, Harvard Business School Working Paper 26-063, March 2026). Traditionally, expertise resides with workers and is hard to transfer because it is tacit; firms therefore “rent” it period by period. Surveillance-enabled AI changes this: by recording how workers do their jobs, firms can codify that expertise into AI systems they own and scale. In the model, the records workers generate while working train an AI that becomes the firm’s future outside option — and a stronger outside option lets the firm push wages down. Surveying employed US workers, the authors find that large majorities hold undocumented knowledge and judgment, and that awareness of AI-training use causes workers to withhold data and raise their reservation wages.
Scenario 1 — selling captured expertise
A firm that owns both the trained model and the underlying data (through work-for-hire plus database and trade-secret protection) could, in principle, license or sell the resulting model or its outputs to other firms without compensating the worker — because the worker holds no property right in their tacit knowledge or in the employer-owned data. This is a plausible extension rather than a documented mass practice; it is the logical endpoint of treating labour as capital.
Scenario 2 — declining to hire or train juniors
The apprenticeship ladder is directly threatened. If a senior worker’s expertise is embedded in an AI agent, the marginal value of hiring and training a junior falls. Emerging work provides early evidence of adverse AI-exposure effects concentrated on early-career workers and warns of “seniority-biased” technological change with long-term harm to the intergenerational transmission of knowledge; related analysis estimates that lost on-the-job learning could shave measurable points off productivity growth.
The full range of consequences for employees
● Wage suppression and lost bargaining power — an improved employer “outside option” lets firms push wages down.
● Erosion of the scarcity value of expertise — AI spreads the best practices of the ablest workers, compressing the skill premium.
● Reduced mobility and lock-in — once expertise is embedded in an employer-proprietary system, the worker cannot take that asset elsewhere (unlike portable skills), and non-competes and IP clauses compound the lock-in. The now-vacated US federal analysis found non-competes bind roughly one in five workers and suppress wages.
● Collapse of the apprenticeship ladder — fewer junior roles, harming entry-level workers and the future senior pipeline.
● Deskilling and automation bias — documented in fields such as radiology and pathology, where reliance on AI advice can override correct independent human judgment, especially under time pressure, and where shrinking caseloads erode the morphological expertise that training depends on.
● “Teach then discard” — the worker becomes a one-time trainer of the AI and is then made redundant.
● Inability to monetise one’s own experience — the value accrues to the firm’s model rather than the worker.
● Information and power asymmetry — workers often cannot see what data or logic governs decisions about them, and trade-secret claims are used to resist data-access requests.
● Harm to older and experienced workers — their key differentiator is neutralised.
● Professional effects — across law, medicine, consulting, software, translation, customer service, and creative work.
● Macro effects — monopsony, concentration of human-capital value in firms rather than workers, and a possible shift in the capital–labour income share.
● Psychological and dignity harms — loss of autonomy, professional judgment, and the craft “opacity” that sustains negotiating leverage.
A counterargument worth stating
In the short run, AI has helped novices and low-skilled workers most, reducing within-firm inequality and improving retention. The radiology case is instructive: a widely publicised 2016 prediction that AI would make radiologists obsolete did not materialise, and demand for the specialty has instead risen. The harm, in other words, is contingent on ownership rules and bargaining power — it is not technologically determined.
Part (C) — Recommendations
For labour unions
● Bargain over data and AI as a mandatory subject. Treat the whole data lifecycle — collection, use, training, licensing, deletion — as a bargaining item, drawing on established workers’ data-rights principles and model collective-agreement clauses.
● Pursue collective — not individual — data ownership. Because individual ownership can backfire, establish workers’ data trusts or collectives that internalise the externality and let workers act as a bloc.
● Adopt the Writers Guild / SAG-AFTRA template. Transferable provisions include: AI output cannot displace credit or reduce minimum staffing; consent and compensation for digital replicas; reserved rights to contest AI training on covered work; and mandatory employer–union meetings to monitor AI use. Adapt the principle to: AI trained on my work product or behaviour requires consent and compensation.
● Demand revenue-sharing or royalties where worker data trains a model, recognising the act of “teaching” automation as compensable work.
● Secure transparency, access, and audit rights beyond the statutory floor — the right to know what data is collected, what logic is used, and to audit the system (the professional-athlete wearable-data precedent is a concrete model of full worker access).
● Protect the experience ladder — negotiate retention of junior and training roles and reskilling obligations.
● Use co-determination where available. The German works-council model gives binding rights: early information about AI planning, a legal presumption that the council may engage an employer-funded external expert when AI is involved, consent rights over AI-based personnel-selection guidelines, and co-determination over systems capable of monitoring performance or behaviour.
For regulators
● Close the purpose-limitation loophole. Expressly prohibit repurposing monitoring, quality, or security data for AI training without a fresh lawful basis, an impact assessment, and worker consultation.
● Mandate transparency and impact assessments. Require Workplace AI Impact Assessments before deployment, following existing model-bill blueprints.
● Guarantee human-in-command. Require meaningful human review, ban fully automated dismissal (as the Platform Work Directive already does), and mandate information and consultation.
● Extend collective rights. Promote sectoral bargaining and collective data rights, building on the AI Act’s worker-notice duty and the Platform Work Directive’s collective-representation provisions.
● Consider a benefit-sharing mechanism — a statutory right for workers to share in value derived from models trained on their work.
● Strengthen worker bargaining power broadly — the most robust lever: restore limits on non-competes, support unionisation, and enable portable benefits.
● Fund reskilling and worker-complementary AI — steer public R&D toward augmentation rather than pure substitution.
● Fill the US gap — enact a federal baseline and expand the bias-audit model beyond bias to cover expertise capture and data reuse.
A staged roadmap
Immediate (0–12 months)
● Unions: table AI and data clauses at the next bargaining round; inventory what monitoring data employers already hold and how it is used; and enforce existing works-council and worker-notice rights now (already required in the EU).
● Regulators: issue guidance clarifying that repurposing monitoring data for AI training is a new purpose requiring a fresh lawful basis, an impact assessment, and consultation.
● Escalation trigger: evidence that any employer is licensing or selling worker-derived models, or a measurable decline in junior hiring in an occupation, should trigger sectoral bargaining and legislative action.
Medium term (1–3 years)
● Establish workers’ data trusts; negotiate revenue-sharing where models are trained on worker data.
● Regulators: legislate Workplace AI Impact Assessments, ban fully automated dismissal, and mandate audit and access rights — extending platform-work protections to all workers, not only platform workers.
● Escalation trigger: if wage data show senior-worker wage erosion attributable to AI capture, move to mandatory benefit-sharing.
Long term (3+ years)
● Sectoral collective agreements on AI; portable benefits; and publicly funded worker-complementary AI research.
● Consider statutory recognition of a collective worker interest in models trained on their labour.
Caveats and confidence
● Much of the strongest economic evidence is a working paper. “Labor as Capital” is explicitly a draft; its conclusions are model- and survey-driven, not causal field evidence on wage effects. Treat them as theoretically rigorous but not yet confirmed at scale. By contrast, the customer-support productivity study is peer-reviewed.
● Scenarios 1 and 2 are largely prospective. No documented case yet exists of a firm openly reselling an individual worker’s captured-expertise model; it is a logical extrapolation. Junior-hiring harm has early but contested evidence.
● The skill-compression finding cuts both ways. In the near term, AI has helped novices most and improved retention; whether it becomes net-harmful depends on ownership and bargaining rules, not the technology alone.
● Human-rights instruments are a weak direct remedy. The authorship rights in the Universal Declaration and the ICESCR protect creative and scientific productions, not tacit job skills, and bind states rather than private employers; their application to expertise capture is aspirational rather than established law.
● Regulatory flux. The EU’s late-2025 “Digital Omnibus” proposal may delay some AI Act deadlines; the US federal non-compete ban is dead; and US federal action is uncertain. Dates and applicability cited are current as of July 2026 but may change.
● Source-quality note. Some data-protection explainers cited are compliance-vendor summaries; the core legal points have been corroborated against primary texts, regulators, and case law. Some supporting empirical details (for example, specific deskilling figures) derive from the wider research literature and are stated in general terms where a single authoritative figure could not be independently confirmed.
Sources
All URLs were verified at the time of writing (22 July 2026).
Primary legal instruments, cases, and official texts
European Union, Regulation (EU) 2024/1689 (Artificial Intelligence Act), EUR-Lex. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
European Commission, Digital Omnibus on AI (proposal amending Regulation (EU) 2024/1689), EUR-Lex. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex:52025PC0836
European Union, Directive (EU) 2024/2831 (Platform Work Directive), EUR-Lex. https://eur-lex.europa.eu/eli/dir/2024/2831/oj
UN Committee on Economic, Social and Cultural Rights, General Comment No. 17 (E/C.12/GC/17) on Article 15(1)(c) of the ICESCR (Refworld). https://www.refworld.org/legal/general/cescr/2006/en/39826
UN Committee on Economic, Social and Cultural Rights, General Comment No. 17 (UN Digital Library record). https://digitallibrary.un.org/record/566430?ln=en
International Covenant on Economic, Social and Cultural Rights (ICESCR), Office of the UN High Commissioner for Human Rights. https://www.ohchr.org/en/instruments-mechanisms/instruments/international-covenant-economic-social-and-cultural-rights
Universal Declaration of Human Rights, United Nations. https://www.un.org/en/about-us/universal-declaration-of-human-rights
Bărbulescu v. Romania (European Court of Human Rights, Grand Chamber, 2017) — overview. https://en.wikipedia.org/wiki/B%C4%83rbulescu_v._Romania
O’Connor v. Ortega, 480 U.S. 709 (1987) (FindLaw). https://caselaw.findlaw.com/court/us-supreme-court/480/709.html
17 U.S.C. § 101 (definitions, including “work made for hire”), Cornell Legal Information Institute. https://www.law.cornell.edu/uscode/text/17/101
New York City Department of Consumer and Worker Protection, Local Law 144 — Automated Employment Decision Tools(official page). https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page
Office of the New York State Comptroller, Enforcement of Local Law 144 — Automated Employment Decision Tools (audit, 2 Dec 2025). https://www.osc.ny.gov/state-agencies/audits/2025/12/02/enforcement-local-law-144-automated-employment-decision-tools
US Federal Trade Commission, FTC Announces Rule Banning Noncompetes (2024; rule subsequently vacated, appeal withdrawn Sept 2025). https://www.ftc.gov/news-events/news/press-releases/2024/04/ftc-announces-rule-banning-noncompetes
Empirical and economic research
Brynjolfsson, Li & Raymond, “Generative AI at Work,” The Quarterly Journal of Economics 140(2), 2025 (Oxford Academic). https://academic.oup.com/qje/article/140/2/889/7990658
Brynjolfsson, Li & Raymond, “Generative AI at Work,” NBER Working Paper 31161. https://www.nber.org/papers/w31161
Cullen, Li & Li, “Labor as Capital: AI and the Ownership of Expertise,” HBS Working Paper 26-063, March 2026 (landing page). https://www.hbs.edu/faculty/Pages/item.aspx?num=68816
Cullen, Li & Li, “Labor as Capital: AI and the Ownership of Expertise” (full PDF, HBS WP 26-063). https://www.hbs.edu/ris/Publication%20Files/26-063_6b332c34-be84-4af4-9d29-1107d22419e9.pdf
“Knowledge Guilds: Sharing the Productivity Gains of AI,” HBS Working Paper 26-064 (PDF). https://www.hbs.edu/ris/Publication%20Files/26-064_6e7fa6b3-14ce-4817-bdf4-436a4778b6f4.pdf
“A theory-based AI automation exposure index: Applying Moravec’s Paradox to the US labor market,” arXiv:2510.13369. https://arxiv.org/pdf/2510.13369
Intellectual property, trade secrets, and work made for hire
Saunders & Golden, “Skill or Secret? — The Line Between Trade Secrets and Employee General Skills and Knowledge,”SSRN. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3399277
DuPage County Bar Association, “An Employer’s Guide to Copyright Law’s Work for Hire Doctrine.” https://www.dcba.org/mpage/vol301217art2
Data protection and privacy
GDPR Local, “GDPR Employee Monitoring: Compliance Considerations.” https://gdprlocal.com/gdpr-employee-monitoring/
gStride, “GDPR-Compliant Employee Monitoring: 2026 Checklist.” https://gstride.ai/blog/gdpr-compliant-employee-monitoring/
Secure Privacy, “GDPR Article 22 and Automated Decision-Making.” https://secureprivacy.ai/blog/gdpr-article-22-automated-decision-making-guide
Human rights and IP commentary
infojustice (American University), “Make copyright compatible with the UN ICESCR.” https://infojustice.org/archives/32035
AI in employment — legal analysis
Lexology, “EU AI Act: High-risk AI systems in employment — practical steps for compliance.” https://www.lexology.com/library/detail.aspx?g=19b69b8c-4616-47f1-b1fd-a4c77cb790c0
EU Artificial Intelligence Act portal, “What the EU AI Act Means for Staffing Businesses.” https://artificialintelligenceact.eu/what-the-act-means-for-staffing-businesses/
Ethics and the worker perspective
OnLabor, “Training Your Replacement, One Keystroke at a Time.” https://onlabor.org/training-your-replacement-one-keystroke-at-a-time/
Employee Responsibilities and Rights Journal (Springer), “The Case for Selective Non-Transparency in AI-Mediated Work.” https://link.springer.com/article/10.1007/s10672-025-09567-z
Union and policy templates
Writers Guild of America, “Artificial Intelligence” (know-your-rights, 2023 MBA provisions). https://www.wga.org/contracts/know-your-rights/artificial-intelligence
Writers Guild of America East, “Summary of the 2023 WGA MBA.” https://www.wgaeast.org/guild-contracts/mba/summary-of-the-2023-wga-mba/
Trades Union Congress, “Artificial Intelligence (Regulation and Employment Rights) Bill.” https://www.tuc.org.uk/research-analysis/reports/artificial-intelligence-regulation-and-employment-rights-bill
Trades Union Congress, “The AI Bill Project.” https://www.tuc.org.uk/research-analysis/reports/ai-bill-project
