Featured image: AI-generated editorial illustration for Bubbling Box. A conceptual scene of human-led technology and community work—not an existing program or actual deployment.

Paid civic work, cleaner water and disaster-ready infrastructure are at the center of a citizen-authored AI proposal. Here is what it would do, what it would cost, and what still needs work.

Disclosure and status: This proposal was developed by Bubbling Box’s Chetuno Ebuwa with AI-assisted drafting. This is an explanation of our own proposal, not independent reporting on legislation sponsored by someone else. The source is discussion draft v0.3-S1, dated September 11, 2026. It is not introduced or enacted legislation, and the programs described below are not accepting applications. No congressional or agency endorsement is claimed. (Draft)

Imagine a neighborhood recovering from a hurricane. A trained crew brings temporary drinking-water equipment, power for essential services and a communications link. Robots inspect places too dangerous for people to enter. Local residents have paid roles in preparation, maintenance and recovery, while qualified professionals remain responsible for safety.

That is an illustration of the system proposed here—not a report of a deployment that has happened.

The starting question is simple: what should technological progress do for the people living alongside it? This proposal’s answer is not just faster software. It is useful public work, reliable services, healthier environments and a way for people to help shape the systems affecting their lives.

Its full name is the American Artificial Intelligence Public Benefit, Rights, and Innovation Act of 2026. For now, it is a citizen-authored discussion draft seeking criticism, technical review and potential congressional sponsorship. Readers should judge it as a proposal, including its costs and unresolved questions—not as a promise of benefits already available. (Draft)

People would remain in command

The draft establishes three levels of human–AI authority. At the first, AI assists with analysis, forecasts and recommendations. At the second, robots or software perform defined tasks within tested boundaries under trained supervision. At the third, accountable human beings approve consequential decisions affecting life, rights or substantially irreversible interventions. (Draft, §§101–106)

A recycling machine sorting material and an official deciding whether to evacuate a neighborhood would not receive the same authority. One installation could contain all three levels, depending on the function. A narrowly preauthorized safety control could still react immediately when waiting for approval would create greater imminent danger. (Draft, §§102, 1003)

The human role could not be symbolic. Supervisors would need the training, time, information and practical authority to disagree or intervene. Critical teams would preserve expertise and exercise safe fallback procedures when AI is unavailable or unreliable. The point is useful automation without allowing an organization to escape responsibility by blaming a machine. (Draft, §§101–106)

A paid place for Americans in the work

The proposed American Civic Development and Resilience Corps would combine voluntary civilian employment with skills development. Ordinary adult eligibility would include U.S. citizens and nationals aged 18 or older. Applicants would not need a college degree or proof that AI caused their unemployment. A separate educational track for 16- and 17-year-olds would permit only lawful, supervised, nonhazardous work. (Draft, §§302–303)

The draft would prioritize qualified unemployed, involuntarily underemployed and low-income applicants for at least 60% of annual new adult subsidized placements nationally, while retaining opportunities for other eligible applicants. An unfilled priority placement could open to others after documented outreach and a 30-day recruitment period. (Draft, §303)

Ordinary adult Corps work would start at a proposed $20-per-hour floor in the first program year, with annual inflation adjustment and higher applicable rates taking precedence. Initial placements would generally last six to twelve months. Required training would be paid, and ordinarily at least one-fifth of paid program time would involve structured instruction, supervised skill practice or assessment. (Draft, §§304–305)

Work could include supervised leak surveys, watershed restoration, repair-and-reuse projects, accessibility assessments, emergency inventory, robotics maintenance and public-service navigation. A beginner might learn water-sampling procedures and eventually seek a utility apprenticeship; completing a course would not itself authorize that person to operate a treatment plant. (Draft, §§304–305)

The distinction matters: useful entry-level work is not permission to put unqualified people into hazardous conditions. The draft also bars replacing existing workers principally to obtain subsidized labor, undermining bargaining agreements or using participants to break a strike. Participation would be voluntary, with assistance for documented barriers such as transportation, childcare and disability access. (Draft, §§305–306)

This is not an unlimited job guarantee. Positions and support would depend on available funding, qualifications and safe delivery capacity. A person could receive a reasoned waiting-list decision rather than an immediate placement. Declining the Corps would not, by itself, terminate unrelated benefits, although earnings could still affect other income-tested programs under their own rules. (Draft, §§303, 305–306)

Water first: preserve it, reuse it, protect it

The water program would support leak repair, treatment and reuse, aquifer stewardship, monitoring, and responsible desalination research and demonstrations. Its planning order is to examine conservation, infrastructure repair, reuse and watershed protection before relying on new supply. That would not require withholding lifesaving water until every conservation project is complete. (Draft, §§501–503)

Desalination would be judged on more than freshwater output. Projects would have to assess energy use, delivered-water cost, intake effects, brine management, marine impacts, reliability and trained staffing. A unit qualified to treat freshwater would not automatically qualify to desalinate seawater. (Draft, §503)

Environmental robotics could support waterway cleanup, pollution sampling, lost fishing-gear recovery and infrastructure inspection. Ecological expertise would remain central. In recycling, the program would distinguish collecting waste from actually returning material to productive use; a pile of sorted plastic would not, on its own, count as a successful recycling outcome. (Draft, §§601–602, 701–702)

Disaster resilience before and after an emergency

The proposal combines permanent defenses with reusable mobile services. Permanent projects could include stormwater improvements, wildfire resilience, safer siting, resilient utilities and scientifically appropriate natural defenses. Mobile equipment could temporarily support drinking water, sanitation, critical power, communications, medical support and shelter when existing systems fail. (Draft, §§801–903)

“Disaster prevention” does not mean promising to stop hurricanes or eliminate drought. It means reducing avoidable exposure, preventable ignition, infrastructure failure and damage, while improving detection and warning. Qualified professionals would retain responsibility for official scientific products and emergency decisions. (Draft, §801)

The draft’s planning portfolio includes up to 30 regional hubs, 120 satellite depots, 400 mobile-service packs and 2,000 robotic mission kits. These are planning quantities requiring engineering, procurement and field review—not equipment already purchased, certified or available for dispatch. (Draft, §§902, 1001; response-cost annex)

Interoperability is part of the design: equipment from different suppliers would need to exchange usable information and connect safely. A connection would not give another machine permission to take command. Equipment would also be tested for degraded conditions, and reusable modules would need maintenance, qualified operators and safe commissioning—not just a place in storage. (Draft, §§903, 1005)

The proposed services cover inland, coastal, rural, urban, Tribal, territorial and remote communities. Deployment targets would account for safe routes and operating limits. There is no promise that a drone can fly through hurricane-force winds or that a damaged road becomes passable because an algorithm finds it on a map. (Draft, §§902, 1002, 1602–1603; implementation annex)

Public service without a political loyalty test

A proposed Federal AI and Public Resilience Commission would administer the program through agreements with existing agencies, rather than automatically absorbing their responsibilities. It would have seven commissioners appointed by the President with Senate confirmation, with no more than three affiliated with the same political party. That design is a proposed safeguard, not proof that neutrality can be guaranteed. (Draft, §§1101–1102)

The more important protection is a duty governing actual decisions: assistance or a paid placement could not be denied, delayed or favored because of political affiliation, voting history, campaign contributions or lawful criticism of an official. Published eligibility criteria, recorded reasons, independent audits and proposed review routes would help make those decisions contestable. (Draft, §§1103–1106, 1404)

The reporting calendar is specific. Operational information would update at least weekly in ordinary conditions and daily during major deployments where safe and practical. Financial information would update monthly. Quarterly reports would go to Congress and the public together, followed by an annual independently audited report. Protected personal and security information would not be released indiscriminately. (Draft, §1107)

The draft does not promise an agency beyond constitutional presidential authority, court review or future legislation. Nor does a direct-application pathway simply erase valid local safety rules. Those legal boundaries and the proposed remedies require specialist scrutiny before introduction. (Draft, §§1102, 1105, 1701–1703)

What the $1.2 trillion proposal means

The ten-year proposed envelope is $1.2 trillion. Within that amount, $450 billion would be supplied through scheduled direct appropriations if enacted, while $750 billion would require later appropriations. An authorization allows Congress to provide funding later; it is not money already available to spend. These figures describe proposed budget authority, not a forecast of cash paid out each year. (Draft, §§1501–1502)

Proposed ten-year mission allocations — not enacted funding

Proposed mission	Ten-year allocation
Water security and desalination	$170 billion
Ecosystems, pollution, recycling and agriculture	$110 billion
Weather observations, forecasting and warnings	$60 billion
Permanent disaster-resilient infrastructure	$200 billion
Mobile infrastructure and disaster robotics	$150 billion
Healthcare and public-health AI	$90 billion
Housing and construction robotics	$70 billion
Education, legal access and civic services	$70 billion
Civic work, workforce development and transition	$160 billion
Scientific discovery, public compute and AI safety	$70 billion
Efficient AI energy and water infrastructure	$20 billion
Commission, independent oversight and evaluation	$18 billion
Program contingency and continuity reserve	$12 billion
Total	$1,200 billion ($1.2 trillion)

Source: discussion draft, Title XV, Schedule B. All amounts are proposed ten-year allocations. Subaccounts are included, not added on top. (Draft)

Within the $160 billion workforce account, $90 billion would support paid civic employment. The remainder covers apprenticeships and training, targeted wage insurance, access supports, community-improvement projects, and delivery and evaluation. The proposal does not establish a universal AI dividend or a new robot tax. (Draft, §§301–302)

The accompanying model estimates approximately 1.06 million full-time-equivalent funded job-years over the decade, assuming full funding and its stated costs. One person employed for one full-time year represents one job-year; that is not a count of permanent jobs, unique participants or verified net job creation. Supervision, training, benefits and equipment consume part of the employment budget. (Draft, workforce annex; budget workbook)

This is a large spending proposal, and it is not presented as paid for. It supplies no dedicated financing package, official congressional cost estimate or modeled debt-service estimate. Speculative AI profits and hoped-for savings are not counted as offsets. Without financing changes, additional outlays could increase borrowing. These are questions for serious budget review, not details to conceal behind enthusiasm. (Draft, §1502; budget workbook)

More than emergency equipment

Healthcare, affordable housing, education, legal access, agriculture, scientific research and shared computing remain part of the framework. The proposed human-review and integrity protections also address consequential AI decisions, knowing misrepresentations and harmful synthetic impersonation, while preserving protected expression. Ordinary model error would not automatically become a violation. (Draft, Titles II, IV, XIII–XIV)

The Commission would not become an unlimited regulator of all AI or take over professional licensing. A project could use a non-AI solution when that solution better meets the need. The objective is a better service, not an AI label on every purchase. (Draft, §§401, 403, 1101, 1401)

A system that would have to learn from its mistakes

The draft calls for 25 competitively selected pilot service areas and independent evidence before expansion. Every mission would define its baseline, costs, conventional alternative, safety requirements and intended result. (Draft, §§403, 1602)

Monthly team reviews, quarterly cross-region reviews and annual development plans would turn worker feedback and public experience into proposals for improvement. At least 3% of each mission’s annual allocation, excluding the separate contingency reserve, would support suitable testing, evaluation and corrective-action verification inside the existing budget. (Draft, §§1605–1607)

Changes to high-consequence systems would require appropriate testing and human authorization. AI could not silently rewrite eligibility criteria or safety limits. Findings of no improvement or insufficient evidence would remain legitimate outcomes. A failed tool should trigger correction or replacement—not automatically strand the people relying on the service. (Draft, §1606)

What happens next—and how to learn more

The immediate task is to improve the proposal through public questions, technical criticism, legal review and congressional conversations. It still needs scrutiny of agency overlap, employment design, equipment assumptions, constitutional questions and financing. An interested congressional office might prefer a separately drafted first-stage measure, but no smaller version has been adopted in this source edition. (Draft, submission and review notes)

A citizen draft is not an introduced bill. Introduction, committee consideration and the rest of the legislative process are separate steps; the House explains that process in its public guide. CBO provides Congress with nonpartisan budget analysis and cost estimates, but this proposal has no official CBO score. (House guide) (CBO overview)

The useful next question is not whether readers must agree with every provision. It is which parts would genuinely improve community life, what evidence would establish that, and what should change before public money is committed.

Read the source documents, check the assumptions and identify specific improvements. That is how an ambitious idea becomes something people can examine rather than simply applaud.

Read the proposal and its sources

Source edition: v0.3-S1, September 11, 2026. The full proposal governs if a summary differs.

1. Complete legislative discussion draft — PDF. Proposed text, implementation assumptions and review notes; not enacted law.
2. Three-page overview — PDF. An introduction for readers and legislative staff.
3. Budget and implementation workbook — Excel. Planning allocations, formulas and assumptions; not an official congressional score.
4. U.S. House: The Legislative Process.
5. Congressional Budget Office: Introduction to CBO.

For readers who wish to ask their own offices about legislative review, the official House representative locator and Senate contact directory provide contact routes. Use your real constituency and your own words; this page is not a petition, fundraising appeal or program-enrollment service.

Written by Chetuno