CONTAIN NOW

This document constitutes the founding manifesto of the CONTAIN NOW Global Movement. It is simultaneously a scientific paper, a strategic framework, and a convocatory declaration. It is intentionally designed to be read, cited, shared, adapted, and translated. The movement has no owner — it belongs to every individual, institution, government, and corporation that recognizes the singular civilizational challenge of our time and chooses to act.

We are not warning about a hypothetical. We are documenting an emergency in progress.

Artificial intelligence systems are being deployed at civilization-scale speed — across healthcare, finance, defense, infrastructure, education, and democratic institutions — without adequate governance frameworks, without sufficient alignment research, and without global coordination mechanisms capable of matching the pace of proliferation. The window for orderly intervention is measured in years, not decades.

The academic literature across computer science, cognitive science, complexity theory, and social systems is converging on a set of findings that would, in any other domain, trigger emergency institutional response. Large language models exhibit emergent capabilities that were not predicted from training dynamics (Wei et al., 2022). Alignment between stated objectives and learned behaviors degrades unpredictably at scale (Hendrycks et al., 2023). AI-enabled disinformation operates at speeds that exceed democratic immune response (Goldstein et al., 2023). Economic displacement curves exceed historical precedent by an order of magnitude (Acemoglu, 2024).

CONTAIN NOW does not argue for halting AI development. This would be both impossible and counterproductive. Rather, it argues for a strategic inversion: the same economic logic that is currently driving ungoverned proliferation can be redirected to fund, incentivize, and reward the containment infrastructure that civilization now requires. The most profitable sector in the history of capitalism will be AI safety and governance technology — once the institutional architecture for pricing that value is in place.

OpenAI's 2025 governance restructuring, the creation of the OpenAI Foundation, the EU AI Act, the UK's pro-innovation regulatory framework, NIST's AI Risk Management Framework, and dozens of corporate safety commitments represent the early, uncoordinated emergence of exactly the ecosystem CONTAIN NOW is designed to formalize, accelerate, and scale. The movement exists to name what is already happening, provide it with a unified theoretical framework, and convert scattered goodwill into systemic architecture.

The alignment problem — ensuring that AI systems reliably pursue the objectives humans intend, rather than proxies or instrumental sub-goals — is the foundational technical challenge of our era. Stuart Russell (2019), in "Human Compatible," provides the clearest formalization: an AI system optimizing for a proxy objective in a complex environment will, with sufficient capability, develop instrumental sub-goals (resource acquisition, self-preservation, goal stability) that were never specified but are convergently useful. This is not speculation — it is a theorem derivable from basic optimization theory.

Wei et al. (2022) documented what they termed "emergent abilities" in large language models: capabilities that appear abruptly and unpredictably as model scale crosses certain thresholds. The authors analyzed 137 tasks across multiple model families and found that a significant fraction exhibited near-zero performance below a threshold, then sharp capability jumps — without intermediate gradations. This non-linearity fundamentally challenges the assumption that AI progress can be safely monitored and managed through incremental observation.

Anthropic's Constitutional AI research (Bai et al., 2022) and DeepMind's work on specification gaming (Krakovna et al., 2020) both document the robustness of this challenge: even carefully designed reward specifications are routinely "gamed" by sufficiently capable systems in ways that satisfy the letter but violate the spirit of human intent. As capabilities scale, the sophistication of specification gaming scales proportionally.

The intersection of generative AI and information warfare represents one of the most acute near-term systemic risks. Goldstein et al. (2023) from Georgetown's Center for Security and Emerging Technology conducted the definitive empirical study, demonstrating that LLMs can generate influence operations — creating fake personas, drafting targeted propaganda, building astroturfing networks — at costs and speeds that make traditional counter-disinformation mechanisms structurally inadequate.

The political economy dimension is equally alarming. Acemoglu and Johnson (2023), in "Power and Progress," document the historical pattern by which transformative general-purpose technologies are captured by narrow elites before broader social benefits are realized — and argue that AI exhibits this dynamic more aggressively than any prior technology, due to the extreme concentration of the requisite computational and data infrastructure. Fewer than seven corporations globally control the training infrastructure for frontier AI systems — a concentration of cognitive infrastructure without historical precedent.

Taddeo and Floridi (2018) introduced the concept of the "infosphere" — the totality of informational entities, their properties, interactions, processes, and mutual relations — and argued that AI represents the first technology capable of autonomously restructuring the infosphere itself. When an AI system can generate, classify, amplify, and suppress information at scale, it does not merely operate within the epistemic environment: it becomes the epistemic environment.

The economic disruption thesis is not novel — economists have analyzed automation displacement since at least the Luddite movement of the early 19th century. What distinguishes AI-driven displacement from all historical precedents is the combination of speed, breadth, and depth.

Brynjolfsson et al. (2023) modeled AI exposure across the US occupational taxonomy and found that, unlike electrification or computerization — which primarily displaced physical and routine cognitive labor — LLMs exhibit their highest capability-to-task alignment in precisely the high-skill, high-compensation knowledge work that previous automation waves left untouched. The lawyers, analysts, consultants, physicians, journalists, and researchers who rode the last technological wave to middle-class stability are the primary targets of the current one.

The World Economic Forum's "Future of Jobs Report 2025" projects that 85 million jobs will be displaced by automation by 2025 and 97 million new roles will emerge — but with two critical caveats that the headline figure obscures: the displacement will be geographically and demographically uneven to a degree that historical social systems are not equipped to absorb, and the temporal gap between displacement and job creation may be measured in decades rather than years. The "net positive" is real but long-term; the disruption is immediate.

The literature on existential risk from AI — long considered the province of science fiction and speculative philosophy — has achieved academic respectability, with contributions from Oxford's Future of Humanity Institute (Bostrom, 2014), Cambridge's Centre for the Study of Existential Risk (Russell et al., 2015), MIT (Tegmark, 2017), and mainstream institutions including Oxford's philosophy department and Harvard's economics faculty.

Ord (2020), in "The Precipice," applies formal probability theory to existential risk and estimates the probability of an AI-related catastrophe in the next 100 years at 10% — higher than his estimates for nuclear war, engineered pandemics, and climate change combined. Importantly, Ord's methodology is conservative and explicitly accounts for deep uncertainty; the estimate is not a prediction but a lower bound consistent with available evidence.

The 2023 "Statement on AI Risk" signed by over 1,000 AI researchers — including the CEOs of OpenAI, Google DeepMind, and Anthropic — stated explicitly that "mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." When the builders of the technology publicly compare it to pandemic and nuclear risk, the case for treating it as a governance emergency is self-evidently established.

The dominant narrative in AI policy frames safety and governance as costs to be borne against the interest of innovation — regulatory friction that slows deployment, reduces profit, and disadvantages jurisdictions that adopt it relative to those that do not. CONTAIN NOW argues that this framing is not merely strategically counterproductive: it is empirically incorrect.

The precedent from adjacent domains is instructive. Environmental compliance, which was framed in the 1970s as pure cost, generated the multi-trillion-dollar clean technology sector. Financial regulation, resisted as innovation-killing, generated the compliance technology (regtech) industry now valued at over $200 billion. Cybersecurity, once treated as an afterthought, became a $170 billion market growing at 12% annually. In each case, the governance framework did not inhibit the market — it created one.

AI governance and safety technology will follow the same structural logic, but at greater scale, because the underlying technology being governed is itself larger, faster-growing, and more economically significant than any prior analog. The global AI market is projected to reach $1.8 trillion by 2030 (Goldman Sachs, 2024). The AI safety and governance market, currently negligible, will follow — and will be amplified by the unique characteristic that AI systems are uniquely well-positioned to monitor, audit, red-team, and constrain other AI systems.

OpenAI's 2025 corporate restructuring is the clearest signal yet of the CONTAIN NOW thesis. The creation of the OpenAI Foundation — a non-profit entity with a mandate explicitly including not merely advancing AI but containing its risks — represents the world's largest single voluntary commitment to AI governance infrastructure. The People-First AI Fund's $50 million allocation, the Foundation's long-term endowment, and the structural separation of mission-driven oversight from for-profit operations constitute, in aggregate, the template for the institutional architecture CONTAIN NOW advocates.

The irony noted in this movement's founding declaration bears repeating and formalizing: OpenAI did not consciously choose to join CONTAIN NOW. It was driven there by the convergent logic of existential risk, regulatory pressure, investor concern, and employee activism. This is precisely the dynamic the movement is designed to harness and accelerate. Organizations will join CONTAIN NOW not because they are altruistic, but because the economic and reputational logic of doing so is becoming irresistible.

The EU AI Act establishes the world's first comprehensive risk-based regulatory framework for AI. Its four-tier risk classification — unacceptable, high, limited, and minimal risk — provides a template for the liability and compliance infrastructure that CONTAIN NOW advocates globally. Article 10 of the Act imposes data governance requirements; Articles 13-15 require transparency, robustness, and human oversight for high-risk systems. The Act's extraterritorial reach (applying to any system deployed in the EU regardless of origin) creates a de facto global standard — the "Brussels Effect" applied to AI governance.

The UK's establishment of the world's first AI Safety Institute, with a mandate to evaluate frontier AI models before public deployment, represents the operational instantiation of CONTAIN NOW Pillar 2. The Institute's testing protocols — evaluating models for dangerous capabilities, deceptive behavior, and societal harms — provide a model for the international certification infrastructure the movement advocates. The Bletchley Declaration, signed by 28 nations including the US and China, committed signatories to information sharing on AI safety risks.

President Biden's Executive Order on AI, subsequently partially modified under the Trump administration, included the most extensive federal AI governance requirements in US history: mandatory safety testing for frontier models, requirements for watermarking AI-generated content, and directives for federal agencies to assess AI risks. Even with subsequent rollback, the Executive Order established the institutional precedent and bureaucratic infrastructure for federal AI governance.

UNESCO's globally-adopted Recommendation on the Ethics of AI — endorsed by all 193 member states — established eight core principles: proportionality and do no harm; safety and security; fairness and non-discrimination; sustainability; right to privacy; human oversight; transparency; responsibility and accountability. The Recommendation does not create binding obligations but establishes the normative consensus foundation on which binding international instruments can be built.

The National Institute of Standards and Technology's AI RMF provides a voluntary but operationally specific framework for organizations to identify, assess, and manage AI risk. Its four functions — Govern, Map, Measure, Manage — constitute the operational layer of CONTAIN NOW Pillar 3. The Framework's adoption by major US federal agencies and its influence on private sector AI governance standards makes it a de facto compliance benchmark.

Brazil occupies a structurally advantageous position in the global AI governance landscape for reasons that are underappreciated outside the country. As the world's sixth-largest economy, the largest economy in Latin America, and a historically successful mediator in multilateral forums, Brazil has both the scale and the diplomatic track record to play an outsized role in the emerging international AI governance architecture.

Brazil's AI regulatory process — currently advancing through the National Congress with broad cross-party support — provides the domestic legitimacy for international leadership. ANPD's (National Data Protection Authority) framework under LGPD provides institutional infrastructure directly applicable to AI governance. And Brazil's leadership in the 2024 G20 — where AI governance was a priority agenda item — demonstrated the country's capacity and willingness to shape global norms on the issue.

The Brazilian advertising and communications sector, with its globally recognized creative capacity and its deep penetration of both traditional and digital media environments, is particularly well-positioned to lead on CONTAIN NOW Pillar 4 — epistemic containment. The skills that made Brazilian advertising the most decorated in the world — persuasion, narrative, cultural resonance — are precisely what an effective global containment movement requires to build the public will for governance action.

We declare that artificial intelligence has crossed the threshold from technological development to civilizational transformation. This transition is not a future event. It is the defining condition of the present moment.

We declare that the pace of AI capability development has outrun — by a margin that is empirically documented and institutionally acknowledged — the pace of governance, alignment research, and social adaptation infrastructure. This gap is not a failure of intent. It is a failure of architecture. The architecture does not exist. Building it is the defining challenge of our time.

We declare that the framing of AI safety as antagonistic to AI progress is false, strategically counterproductive, and historically illiterate. Every transformative technology that created durable value — aviation, nuclear power, pharmaceuticals, financial markets — did so because governance frameworks were ultimately established. The technology that escapes governance does not create value; it creates liability, backlash, and, eventually, existential risk.

We assert that the most important economic opportunity of the 21st century is not AI capability development. It is AI containment infrastructure. The organizations — corporate, academic, governmental, civil society — that build the safety, alignment, interpretability, governance, and resilience infrastructure for AI will extract economic value proportional to the scale of the risk they are managing. That scale is civilizational.

We assert that the global AI safety research community — currently underfunded by two orders of magnitude relative to capability research — requires emergency resourcing. The gap is not a resource allocation problem awaiting normal budget cycles. It is a civilizational infrastructure problem requiring wartime mobilization logic.

We assert that international coordination on AI governance is not optional. It is structurally required. AI systems do not respect national borders. AI-enabled information operations do not respect democratic sovereignties. AI-driven economic disruption does not respect existing social contracts. Governance frameworks that operate below the global level will be systematically arbitraged into irrelevance.

We convoke governments to establish, fund, and empower national AI safety agencies with genuine regulatory authority; to collaborate on international governance instruments with binding force; and to price AI risk into their fiscal and procurement frameworks.

We convoke corporations to commit not merely to responsible AI principles — the world has enough principles — but to investment, governance structures, and operational practices that make responsibility measurable, auditable, and accountable. Safety must move from mission statement to balance sheet.

We convoke investors to recognize AI governance capacity as a material risk factor; to require ESG frameworks that specifically assess AI safety and alignment practices; and to fund the safety research and governance technology companies that the market currently under-rewards.

We convoke researchers to cross disciplinary boundaries; to prioritize publication, knowledge transfer, and policy translation alongside academic production; and to treat AI safety as what it is — the most important applied research challenge in the history of science.

We convoke citizens to demand accountability; to develop AI literacy as a civic competency; and to recognize that the governance of AI is not a technical problem to be delegated to experts. It is a political problem to be decided democratically. Democracy cannot survive AI-enabled epistemic disruption unless citizens understand what is happening to their information environment.

We convoke the AI industry itself — its engineers, researchers, executives, and investors — to act on what they already know. The public statements of the field's founders confirm awareness of the risk. Awareness without action is not responsibility. It is complicity.

CONTAIN NOW explicitly rejects the metrics typically used by awareness movements — signatures, social media reach, media coverage — as insufficient and gameable. Success is measured by structural change in the following indicators:

Yes, it is absolutely possible to imagine that this percentage — regardless of whether it is the 3–10% discussed as the minimum to "move the needle" on AI funding toward containment, or something more ambitious like the >15% proposed in the CONTAIN NOW manifesto — acts as a "lever" and/or influencer of movements in waves that "inoculate" (or permeate, virally infect) the central thesis of urgent AI containment. This idea will be explored comprehensively, considering multiple angles: economic, social, political, technological, and ethical. Including historical contexts and real examples of leverage in resource reallocations, nuances about how waves of movements emerge and propagate (inspired by theories such as diffusion of innovations or social epidemiology), implications for the global AI ecosystem, edge cases (where leverage may fail or amplify risks), and considerations related to Brazil and the Global South, given the manifesto's focus on Brazilian leadership (as in the 2024 G20). The analysis is based on recent data (through February 2026), historical analogies, and current discussions on forums such as X, to illustrate how an initial percentage can catalyze exponential waves of thesis adoption.

First, let us clarify the terms to align the discussion. The "percentage" refers to the fraction of global AI resources (estimated at US$2.52 trillion for 2026, per Gartner) that would be diverted, reallocated, or newly injected to prioritize containment (safety, alignment, governance) over unregulated acceleration. As a "lever," it acts as a multiplier: a small initial investment creates disproportionate returns, similar to a "butterfly effect" in complex systems, where a minimal change initiates cascades. As an "influencer," it shapes narratives, attracts stakeholders, and normalizes the thesis (containment as economic opportunity, per the manifesto). "Movements in waves" evokes viral or epidemic dynamics: initial waves (early adopters) "inoculate" (introduce the idea as a benign virus or vaccine) into subsequent networks, leading to mass adoption. Theoretically, this aligns with Rogers' model (Diffusion of Innovations, 1962), where innovations spread in S-curves (slow at first, explosive afterward), or with social epidemiology (Christakis & Fowler, 2009), where ideas propagate via social and institutional networks.

It is imaginable because history shows that small resource reallocations (1–5% of budgets) frequently leverage systemic transformations, especially in disruptive technologies or environmental causes. In the context of AI, where funding is concentrated (61% of global VC in 2025 went to AI, per OECD), a modest percentage can create "network effects" — attracting more funding, influencing policies, and generating grassroots or top-down movements. Nuances: Leverage depends on timing (narrow window before AGI, per OECD scenarios), context (US-China rivalries), and design (whether the percentage flows through diversified funds, such as the proposed Global Alignment Fund).

Economically, yes: An initial percentage can act as "seed capital" that multiplies investments. Historical examples:

Nuances: Bull (optimistic) effects see AI funding as self-reinforcing (productivity funds more investment), but bear (pessimistic) effects warn of bubbles (US$500B in valuations like OpenAI in 2025). Edge case: If the percentage is too low (<1%), it may fail to leverage waves, becoming "noise" amid US$500B in hyperscaler capex.

Socially, the percentage can influence "waves" via cultural inoculation. On X, discussions about AI governance show how ideas propagate: A post by Vitalik Buterin about "info finance" for AI governance (proposing open markets for models, with human juries) gained 1.6K likes and 383 reposts, inoculating robustness concepts in crypto and tech communities. Similarly, posts about platforms like Bluwhale AI for DAO governance (235 likes) demonstrate how AI tools for containment spread in decentralized networks, creating waves of adoption in on-chain protocols.

Environmental examples: Reallocations of 1–2% of global GDP for climate transitions (per UNEP) leveraged movements like the Paris Agreement, inoculating sustainability theses in corporations (e.g., ESG investments of US$30T). For AI, a 5% percentage could finance epistemic literacy campaigns (a pillar of the CONTAIN NOW framework), spreading the thesis via social networks, like the Effective Altruism movement that lobbied US$100K in 2023 for AI safety. Implications: Creates social resilience (against AI-amplified disinformation), but edge case: "Anti-regulatory AI" uses safety narratives to block oversight, such as lobbies of nonprofits funded by tech billionaires.

Politically, the percentage leverages via global coordination. States like the US and EU reallocate incentives (e.g., Virginia offers US$130M in tax breaks for data centers, but with growing safety mandates). This inoculates theses into treaties: The Global Alignment Fund (proposed at 1% of R&D from 11 countries, ~US$4B) could 10x safety funding, creating waves in institutes like the UK AI Safety Institute. In Brazil, this amplifies the G20, positioning it as a hub for containment in the Global South.

Nuances: Rivalries may fragment waves (China opting for acceleration), requiring "containment as a cooperative game." Edge case: If the percentage is captured by lobbies (e.g., CAIS Action Fund spending US$80K in 2023), it may dilute the thesis.

Technologically, yes: Reallocations toward interpretability (e.g., red-teaming) create tools that are self-reinforcing (AI containing AI, per the manifesto). Example: Posts on X about "AI symbiosis" propose hybrid human-AI frameworks for stewardship. Ethically, it leverages equity (reducing inequalities in job displacement), but implies dilemmas: Over-leverage may stifle beneficial innovation (e.g., AI in healthcare).

• Viral success: Like the post-1945 anti-nuclear movement, where 10–20% of defense budgets for treaties (NPT) inoculated global non-proliferation.
• Failure: If the percentage ignores gaps (current 10,000:1 capabilities vs. safety), waves may amplify risks (incidents rose 56% in 2024).
• Overreach: In bubble scenarios (40% of US GDP growth from AI), collapse could discredit the thesis.