Foundation model developers · AI deployers · Reinsurers · Lloyd's and Munich Re · Property and casualty insurers · Congress · Illinois · State legislatures · European Union
Companion to Who Writes AI Insurance Coverage Conditions — Procurement, Enforcement Decrees, Partner Terms and the Two-Axis Condition · Agentic Foreseeability and the Modern Duty of Care for AI Agents · Who Sets the Agentic Standard of Care — Common Law, State Legislation and Federal Coordination
Foresight-to-Duty Conversion Series Introduction: Meta Foresight-to-Duty Conversion Across Privacy, Child Safety and Scam Ads · Meta and Bank of America Foresight-to-Duty Conversion: Revenue-Funded Versus Cost-Funded Deferral · Consent Order Remedies at Meta, Boeing, Starling and Wells Fargo: Verified Correction Versus Recurrence
Executive Summary
One defect in one widely deployed model can surface at many insureds at once, and no carrier has published capacity for that loss. Gallagher Re’s March 2026 white paper said foundation-model failures could cause linked losses across sectors and accumulate across ceded portfolios in ways existing models cannot capture. The Geneva Association’s 2025 assessment found generative AI failing three of the nine standard insurability criteria, and Munich Re’s enterprise factsheet states limits of up to $100 million across similar-use-case models. Above that figure the loss falls on the firm and the people its agent harmed.
Thesis: Available coverage leaves substantial losses with firms and injured parties, and the remainder is a public policy choice. Insurability has two bottlenecks: credible evidence about controllable losses and sufficient capital for correlated losses. The present paper takes the second. Public policy must decide which risks private markets can absorb, which require pooled funding and what corrective duties should accompany any public support, and three precedents show the choice made before and after a salient loss.
What distinguishes the MindCast approach: Who Writes the AI Coverage Condition?showed how buyers, attorneys general and partners supply the evidence carriers underwrite on. The paper treats the allocation of residual loss as a choice with its own game among developers, carriers, reinsurers and public actors. The paper tests the nuclear, vaccine and terrorism structures against the same correlated-loss scenario. One simulated result carries the argument: state legislatures choosing on their own payoffs prefer financial responsibility to a liability shield by five to one, and Congress introduces a public layer in fewer than one case in four without a loss and in more than three of four after one.
Game mutation is the part of the simulation that matters most here. One qualifying loss changes every player’s payoff at once: Congress moves from waiting to drafting, reinsurers move from measuring accumulation to excluding it and developers that opposed a pool begin to want one. A published aggregation trigger mutates the game in the other direction, because foundation-model failure then becomes a named peril carriers can write rather than one they refuse. A liability shield earned by publication is the third mutation, since it removes the exposure from the game without reducing the harm.
MindCast AI applies Predictive Behavioral Economics + Dynamic Game Theory through MindCast AI Proprietary Cognitive Digital Twin Foresight Simulations (MP CDT FS). Behavioral Economics supplies the decision rules. Game Theory supplies the payoff structure. Predictive simulations emerge from the combination.
What the paper does: The paper separates four loss categories and shows which two the capital bottleneck governs. The paper sets out the stated limits of the affirmative market and the reinsurer’s aggregation choice. The paper then compares three precedents in which public policy addressed an exposure private markets would not absorb, sets out the options they support and names who chooses among them. The paper closes with competing explanations with falsifiers, five Simulation Predictions and a Risk Mitigation Layer.
Simulation Prediction highlights: The forecast set holds five MindCast Foresight Simulation Predictions, three primary and two secondary, stated in full in the Simulation Predictions section. The two a reader can test soonest are a single carrier publicly confirming a bound AI liability policy of $100 million or more for one insured by the end of 2027 (27-48%) and a reinsurer or Lloyd’s publishing an aggregation trigger or sublimit for foundation-model failure by the same date (16-56%). The simulation favors congressional introduction of a pool or backstop bill after a qualifying loss.
🏛️ Policymakers: Decide first which exposures the market absorbs, which need a pool and which need public support, and attach a corrective duty to any public layer. The options can be drafted before a loss and are drafted on worse terms after one.
💼 Executives: Correlated loss from the model you deploy stays with you above the carrier’s limit. Size a reserve against the scenario the company can survive, and ask the model vendor what it would absorb.
⚖️ Counsel: A shield earned by publishing a safety protocol removes the exposure instead of covering it and moves the residual loss to the injured party alone. Each precedent kept a route to tort, so public support need not be drafted as preemption.
📊 Investors: Until a pool or public layer exists, a retained-loss scenario for correlated failure belongs in diligence for every AI deployer and developer in the portfolio.
I Four Loss Categories and the Capital Bottleneck
AI loss falls into four categories that insurance treats differently, and the coverage-condition paper develops the first two. Ordinary deployment loss is an identifiable error within a specified system and within policy limits, which carriers write today on conditions because frequency is observable and the maximum loss is bounded. The coverage dispute is a loss that sits inside a policy and goes unpaid over an exclusion or a notice clause, which is a wording problem rather than a capacity problem. The present paper owns the third and fourth categories.
The third is correlated loss from one widely deployed model, where one defect can surface at many insureds at once and accumulate beyond available coverage. The June survey found foundation-model concentration fails two insurability criteria, loss independence and economic feasibility, and treats maximum possible loss as effectively unbounded because the base of insureds using one model keeps growing.
The fourth is the compensation gap: losses above limits or outside any coverage, which fall on the firm and the injured party. The gap can arise from any of the first three categories, and it is the category public policy must decide about. A public layer is a hypothesis for the third and fourth categories only, and nothing in the research establishes that one is inevitable.
Two bottlenecks run across the categories. The first is evidence about which losses a deployer can control, and the coverage-condition paper shows how verification, litigation data and platform telemetry supply it.
The second bottleneck is capital to absorb a correlated loss. An audit supplies information without supplying claims-paying capital, which is why the Geneva Association’s 2025 assessment found generative AI failing three of Baruch Berliner’s nine insurability criteria: maximum possible loss, loss frequency and information asymmetry. The three failures bind hardest on correlated loss and bind less on ordinary deployment loss as evidence accumulates by any path.
🏛️ Policymakers: An audit mandate addresses the evidence problem and leaves the capital problem untouched, so a pool or public layer is the only instrument that reaches the third and fourth categories.
Takeaway: Correlated loss fails the insurability criteria that ordinary loss passes, and the compensation gap above the private layers is the category public policy must decide about.
II The Affirmative Market States Its Limits
Affirmative AI cover exists, and its stated limits describe the capital question. Armilla’s Lloyd’s-backed policy writes up to $25 million per organization against model errors and agent failures. Testudo writes up to $9.25 million against hallucinations and property damage, and Harbor insures the residual of the agents it rates with its own capital at undisclosed limits. The coverage-condition paper describes what each product underwrites on, and the question here is how much any of them can absorb.
Munich Re’s enterprise product is a separate offer from the Mosaic program. The carrier’s corporate factsheet states “Limits of up to USD 50m per model and USD 100m per models with similar use cases.” The cover extends to own financial loss, legal liability and contractual indemnities to clients. The factsheet describes a premium model of “minimum deposit and variable rate” that can run per output as exposure grows.
Munich Re’s separate provider factsheet caps its standard cover at $5 million. Advertised limits do not establish what buyers obtain, and no bound placement at those limits has been publicly reported.
Reinsurers face the choice that decides whether those limits grow. Gallagher Re’s March 2026 white paper said foundation-model failures could accumulate across ceded portfolios in ways existing models cannot capture. A reinsurer can treat foundation-model failure like a named storm with an aggregation trigger and a sublimit, or exclude it and require standalone cover. The named-storm path pays premium on measured accumulation that exclusion forgoes, and the June 2026 survey recommends a Lloyd’s-style market study on AI aggregation triggers of the kind that resolved silent cyber in 2019 and 2020.
📊 Investors: Advertised limits do not establish what buyers obtain. A confirmed bound placement at $100 million for one insured is the first observable that capacity exists at the scale a correlated loss requires.
Takeaway: The affirmative market states limits in the tens of millions per model and $100 million across similar models, and the reinsurer’s aggregation choice decides whether those limits grow or close.
III Three Precedents Support Several Policy Options
Nuclear power reached the same point in 1957, and Congress acted before any accident. The Price-Anderson Act combined three elements: primary insurance per site at the maximum commercially available, now $500 million; a retrospective industry assessment of about $158 million per reactor per incident, which forms a secondary layer near $15 billion across 95 reactors; and congressional consideration of losses above both. Three Mile Island came in 1979, twenty-two years after the structure.
Vaccines reached it in 1986, and Congress acted after manufacturers began leaving the market under tort exposure. The National Vaccine Injury Compensation Program uses a different instrument: an excise tax on each dose funds no-fault compensation under a table of covered injuries, with a route back to tort for claimants who reject the award. The program substitutes a funded compensation system for litigation rather than backing an insurance market.
Terrorism reached it in 2002, and Congress acted on November 26, fourteen months after the attacks. The Terrorism Risk Insurance Act shares qualifying losses between insurers and the federal government above a deductible, and the program now extends to certified cyberattacks. Critics note that the sharing formula operates in practice as a subsidy, and the CSIS analysis prefers the nuclear structure because “Federal capacity should sit above industry capital rather than in place of it.”
The three precedents support different options rather than one design. The nuclear model backs a private market with an industry assessment and keeps industry capital at risk. The vaccine model replaces litigation with funded compensation for a defined injury table, and the terrorism model shares losses above a threshold without an industry layer.
The triggers differ as well. Two of the three followed a market exit or an attack, and one preceded any loss, so anticipation and crisis both appear as routes to a public layer.
⚖️ Counsel: Each precedent kept a route to tort above or outside the structure, so public support for AI losses need not be drafted as preemption and the two should be argued separately.
Takeaway: The precedents offer a backed market, a funded compensation system and a loss-sharing formula, and they show public policy acting both before and after a salient loss.
IV Policy Options and Who Chooses Among Them
The precedents and the four loss categories yield options that the simulation treats as competing rather than sequential. Market expansion: affirmative capacity grows on evidence from verification, litigation data or platform telemetry, reinsurers set aggregation limits as they do for named storms and ordinary deployment losses become an ordinary line. Compulsory financial responsibility: a state requires deployers above a threshold to show cover, a bond or qualifying assets, which binds only where the market writes it.
Two more options address correlated loss directly. Industry pooling: developers fund a layer in proportion to deployment, drawn only for correlated losses. Public support: a backstop, a funded compensation system or a loss-sharing formula above the private layers.
The opposite design has also been proposed. OpenAI backed an Illinois bill, SB 3444, that would shield frontier developers from catastrophic-harm liability if they published a safety protocol before release. Anthropic opposed it, and the bill’s sponsor said sweeping relief was unlikely to survive. A shield earned by publication pays for files rather than for correction and moves the residual loss to the injured party alone.
The June survey ordered the candidate structures for correlated loss by political burden. Provider indemnification comes first and cedent diversification second; captives and risk retention groups, reinsurance pools and catastrophe bonds follow; a government backstop comes last. The authors expect the earliest visible steps to be aggregation sublimits, provider-event exclusions and captive experimentation, and they recommend a Lloyd’s-style market study on AI aggregation triggers of the kind that resolved silent cyber in 2019 and 2020.
Who chooses is the game, and three actors hold the capital moves. Reinsurers choose between an aggregation trigger and exclusion. Developers hold the evaluation data and the correlated exposure, and their positions on funding a pool are unknown; the reported disagreement between OpenAI and Anthropic concerned liability relief rather than pooling. Congress can authorize a federal mechanism for correlated AI losses, and the institutions paper found Congress slowest absent crisis.
Game theory supplies the payoff structure and behavioral economics the decision rules. A reinsurer that names the trigger collects premium on accumulation that an excluding competitor forgoes, and a developer that funds a pool buys a seat in designing the assessment it would otherwise receive. A state that enacts a shield removes its developers’ exposure and the injured party’s remedy at once. The trigger question is behavioral: two of three precedents moved after a loss and one moved before, and the salient AI loss has not yet occurred.
The simulation plays twenty Cognitive Digital Twins (CDTs) against each other, and predictive behavior emerges from the combination of those payoffs and decision rules. In its modal outcome the performance-guarantee writer binds on provider commitments and reinsurers publish an aggregation limit more often than they exclude. Developers back a pool in one case of four, state legislatures prefer financial responsibility to a shield by five to one while preferring neither in half the cases, and Congress waits unless a loss occurs.
🏛️ Policymakers: The public-layer options can be drafted before a loss and costed against a correlated-loss scenario, and a committee that drafts them first sets the retention and the assessment rather than inheriting them.
Takeaway: The capital options compete, and the forecast turns on whether a reinsurer names the trigger or a loss reaches Congress first.
V Four Explanations Compete for Who Absorbs Correlated Loss
Four readings compete for how correlated loss resolves. Each has its own falsifier, and the Simulation Predictions test them. The coverage-condition paper carries the three readings that concern demand and the upstream chain.
Hypothesis A · Market expansion: Affirmative capacity for ordinary deployment loss grows on evidence from verification, litigation data or platform telemetry. Observables, any two of three by 2028: a publicly confirmed bound placement of $100 million or more per insured with a single carrier; a fall in the excluded share of stated coverage positions below 15 percent in a successor to the June 2026 survey; a reinsurer publishing an aggregation limit for foundation-model failure. Falsifier: fewer than two observables by December 31, 2028.
Hypothesis B · Shield for disclosure: A state enacts liability relief for developers that publish safety protocols, which removes the exposure instead of covering it. Falsifier: no state enacts a catastrophic-harm shield by 2028.
Hypothesis C · Anticipatory public layer: Congress enacts a pool, backstop or compensation system before any correlated loss, as it did for nuclear power. Falsifier: no such enactment by 2028 absent a qualifying loss.
Hypothesis D · Crisis public layer: A qualifying loss is one foundation-model failure producing claims against three or more unrelated insureds that together exceed $1 billion. Given a qualifying loss, a member of Congress introduces a pool, backstop or compensation bill within twelve months of its first public report. Falsifier: a qualifying loss occurs and no such bill is introduced within twelve months. Without a qualifying loss the hypothesis stays unresolved.
The forecasts rest on Hypothesis A as the route that makes any public layer smaller, with C and D as the competing routes to that layer. Hypothesis B is the main threat to correction, because a shield earned by publication resolves the exposure without reducing the harm.
📊 Investors: Under Hypothesis A, correlated loss stays with deployers and injured parties through 2028, so a retained-loss scenario belongs in diligence until a pool or public layer exists.
Takeaway: The evidence favors a market that grows on evidence while the public layer waits for a loss, and the forecasts test whether a shield arrives first.
VI MindCast Simulation Predictions
The forecast set holds five Simulation Predictions: three Primary Simulation Predictions (P) and two Secondary Simulation Predictions (S). The secondary entries are conditional and print as P(A given B). Each entry names an actor, an outcome, a window and a falsifier. Each resolves against a named source.
P-1 · A single carrier or syndicate publicly confirms a bound AI liability policy of $100 million or more for one insured by December 31, 2027 (27-48%)
Advertised limits do not count, including Munich Re’s stated $100 million across similar-use-case models. Layered placements across multiple carriers do not count either. Mechanism: the performance-guarantee writer binds on provider commitments and reinsurer aggregation terms. Falsifier: no publicly confirmed bound placement at that limit by the deadline. Source: carrier and broker announcements and Lloyd’s market notices.
P-2 · A reinsurer or Lloyd’s publishes an aggregation trigger or sublimit for foundation-model failure in a market bulletin or treaty wording by December 31, 2027 (16-56%)
Mechanism: the named-storm path pays premium on measured accumulation that exclusion forgoes. Falsifier: no published trigger or sublimit by the deadline. Source: Lloyd’s market bulletins and reinsurer publications.
P-3 · A state enacts a liability shield for frontier developers conditioned on publishing a safety protocol by December 31, 2027 (2-34%)
Mechanism: legislatures choosing on their own payoffs prefer financial responsibility or the status quo to a shield. Falsifier: no such enactment by the deadline. Source: state session laws.
S-1 · Given no qualifying loss through December 31, 2027, a member of Congress introduces a bill creating a federal pool or backstop or compensation system for correlated AI losses by that date (3-24% given no loss)
A qualifying loss is one foundation-model failure producing claims against three or more unrelated insureds that together exceed $1 billion. Mechanism: an anticipatory bill needs a European liability element and a developer pool position together. Falsifier: no such bill by the deadline absent a qualifying loss. Source: Congress.gov.
S-2 · Given a qualifying loss, a member of Congress introduces such a bill within twelve months of its first public report (75-96% given a loss)
Mechanism: crisis response, as with terrorism in 2002. Falsifier: a qualifying loss occurs and no such bill follows within twelve months. Source: Congress.gov and the loss reports.
Validated outcomes to date: none. No entry has reached its deadline, so the forecast set shows no hit and no miss.
One question remains an open hypothesis without a band: whether a cloud platform publishes limits and perils for the AI cover it distributes.
Takeaway: The bound placement and the aggregation trigger are the nearest-term tests and the shield is a minority route. The simulation favors congressional introduction after a qualifying loss and seldom before one.
VII Risk Mitigation Layer
Each entry states exposure in a unit the stakeholder controls, with severity and probability as separate axes and the band as the probability. Actions are unilateral, with an owner and a deadline tied to the forecast checkpoints. The layer sits as a standalone section so that each audience reads its own entries together; the trade-off is that a reader of one Simulation Prediction must turn here for its actions. Entries for the six audiences on the demand side, from deployers to attorneys general, sit in the coverage-condition paper.
Foundation model developers · P-3 (2-34%), S-1 (3-24% given no loss), S-2 (75-96% given a loss): Severity: high. Exposure: a public layer drafted after a loss on terms set by others, measured in the retention and assessment the statute assigns.
Action 1: share incident data with carriers or a database in a form that supports underwriting, through lawful and confidential channels that protect personal information and commercially sensitive material. Owner: head of safety. Deadline: before the 2027 state sessions open.
Action 2: contract audits that meet Illinois’s standard a year before it binds, so the verification layer exists when a carrier asks for it. Owner: general counsel. Deadline: January 1, 2027.
Residual: a shield-for-disclosure bill, if enacted elsewhere, resets the terms of every later negotiation.
Carriers and reinsurers · P-1 (27-48%), P-2 (16-56%): Severity: moderate. Exposure: a market written by competitors who collected data while the carrier excluded, measured in renewal cycles before re-entry at competitive terms.
Action 1: write conditions with a material-change trigger, a required response and a consequence that preserves third-party payment rather than a blanket exclusion. Owner: product underwriting lead. Deadline: the next form filing.
Action 2: set an aggregation sublimit for foundation-model failure and name the trigger, rather than excluding the peril outright. Owner: chief actuary. Deadline: the January 2027 treaty renewal.
Residual: correlated losses can accumulate beyond the carrier’s available capacity whatever the conditions.
Federal committee staff · S-1 (3-24% given no loss), S-2 (75-96% given a loss): Severity: moderate. Exposure: a public layer drafted in the week after a loss, measured in the share of the loss the public layer absorbs.
Action 1: draft the options now, with the nuclear, vaccine and terrorism structures each costed against a correlated-loss scenario. Owner: committee staff. Deadline: before the 120th Congress organizes.
Action 2: authorize the incident database as a standalone measure that needs no public layer to pass. Owner: committee staff. Deadline: same.
Residual: a drafted option still waits for the vote that enacts it.
Takeaway: Every action builds capital or a drafted option before a loss would otherwise build them on worse terms.
Conclusion
Correlated loss from one model is the exposure the 2026 exclusions left in place and the affirmative market has not absorbed. Munich Re states $100 million across similar models, no carrier has confirmed a bound placement at that figure and no reinsurer has published an aggregation trigger. Above the private layers the loss falls on firms and injured parties.
Public policy decides the remainder. Nuclear power got a backed market before any accident, vaccines got funded compensation after manufacturers left and terrorism got a loss-sharing formula fourteen months after the attacks. The precedents offer three instruments and two triggers, and the simulation expects the crisis trigger to arrive first.
The shield is the alternative that removes the exposure without reducing the harm. A state that enacts one pays for files, and the simulation finds legislatures choosing on their own payoffs seldom do so.
The series now runs from measurement through institutions and evidence to capital. Installment VI asked who writes the condition that carries the reassessment duty, and Installment VII asks who absorbs the loss when the condition is not enough. The answer today is whoever holds the risk when the exclusion attaches, until Congress or a reinsurer changes the rules.
What to Watch
The first checkpoint falls on January 1, 2027, when the reinsurance treaty year renews. State legislatures can introduce shield, pool or database bills through spring 2027, and the 120th Congress organizes in January. The dominant fork is whether a correlated AI loss above $1 billion in aggregated claims arrives before any legislature builds the preconditions.
Each signal maps to an entry. A confirmed single-carrier $100 million placement resolves P-1 and a published aggregation trigger resolves P-2. A shield statute resolves P-3, and a federal pool or backstop bill resolves S-1 or S-2 depending on whether a qualifying loss preceded it.
Working With MindCast
Published analysis can describe the precedents in general. Published analysis cannot tell a specific developer what a pool assessment would cost it, tell a reinsurer which aggregation trigger its cedents would accept or tell a committee which structure absorbs a given loss at the least public cost. The series introduction sets out where the answers sit: deployment footprints, ceded portfolios and the loss scenarios each structure must survive.
MindCast builds Cognitive Digital Twins (CDTs) of carriers and reinsurers, of developers, and of legislatures and committees. The engagement simulates the nuclear, vaccine and terrorism structures and the shield against the same correlated-loss scenarios.
For developers, the output shows what each pool design would assess and which design the other developers would join. For reinsurers, it shows which trigger produces premium rather than exclusion. For committee staff, it costs each public structure against a correlated-loss scenario before a loss sets the terms.
MindCast offers a complimentary pilot built from public filings and backtested on a known case before any retainer. No engagement proceeds without a named decision to resolve. Contact mcai@mindcast-ai.com or visit www.mindcast-ai-simulation.com to schedule the diagnostic session.
Sources
MindCast Works
Who Writes the AI Coverage Condition? Procurement, Enforcement Decrees, Partner Terms and the Two-Axis Condition. Carries the evidence bottleneck, the demand routes and the two-axis condition.
Who Sets the Agentic Standard of Care? Common Law, State Legislation and Federal Coordination. Finds insurers closest to the deployment clock and names the risks carriers will not write.
Agentic Foreseeability and the Modern Duty of Care for AI Agents. Sets the reassessment standard and the per-action insurance design.
Foresight-to-Duty Conversion in AI Governance, Series Introduction. Sets the series question and the commissioning offer.
Consent Order Remedies at Meta, Boeing, Starling and Wells Fargo — Verified Correction Versus Recurrence. Shows why disclosure-conditioned relief produces documentation rather than correction.
Primary Sources
Price-Anderson Act: Nuclear Power Industry Liability Limits and Compensation, Congressional Research Service (2025). Supplies the three-element nuclear structure and its figures.
Armilla AI Raises Lloyd’s-Backed Coverage to $25M as Traditional Insurers Retreat from AI Risk, Armilla (2026). Carrier statement of limit, covered failures and certification condition.
AI Insurance for Corporates, Munich Re factsheet (2025). States limits of up to $50 million per model and $100 million across similar-use-case models and the per-output premium model.
AI Insurance for Providers, Munich Re factsheet (2025). States the $5 million standard-cover cap and per-use-case sublimits.
Munich Re Insure AI: Frequently Asked Questions, Munich Re. Carrier statements on trigger, limits and premium basis.
Secondary Sources
Gregory C. Allen, The Insurance Industry’s Retreat from AI Threatens to Slow Innovation and Adoption, Center for Strategic and International Studies (September 4, 2026). Supplies the exclusion figures, the insurability analysis and the database and backstop recommendations.
Alex Leung, Rex Zhang, Ervin Ling, Kentaroh Toyoda and SiewMei Loh, The Insurability Frontier of AI Risk: Mapping Threats to Affirmative Coverage, Silent Exposures, and Exclusions, AIFT (June 2026). Supplies the 26-product survey, the four-tier frontier and the ordered candidate structures for correlated loss.
Gallagher Re Identifies Systemic Risk From AI Model Failures, Insurance Business (March 25, 2026). Reports the reinsurance broker’s white paper on accumulation across ceded portfolios.
Munich Re aiSure via Mosaic: Parametric AI Error Cover, Actuary.info (2026). Describes the parametric trigger, pre-bind validation and limit.
AI Exclusions Are Spreading. Harbor Just Raised $10 Million to Insure the Gap, Insurance Business (October 19, 2026). Reports Harbor’s agent-level rating and re-rating on change.
AI Liability Insurance Market Map 2026: Every Carrier, Product and Capacity, Agent Insured (August 2026). Reports specialty products and limits as stated by the desk, unverified with carriers.
OpenAI Backs Illinois Bill Shielding AI Firms From Mass Casualty Liability, PPC Land (April 2026). Describes SB 3444 and the sponsor’s statement.
