Any assessment of a company's legal exposure should begin with a principle rather than a tally. A compliance mandate is not a price list. A firm that treats each fine as a cost to be weighed against the revenue its conduct generates has adopted a maxim, and that maxim fails the test of universalization: if every technology company adopted it, the rules that protect users' autonomy would become a mere schedule of fees. With that standard in view, the material reported between early and mid-November 2026 yields a clear finding. Meta faces penalties and suits on many fronts at once, including privacy, child safety, elections, AI training data and frontier AI development itself. Their financial weight differs by orders of magnitude, and most of them are contested.
The argument that follows has three parts. First, the exposure that matters financially is concentrated in two places. Second, the most heavily reported penalties are comparatively small, and their significance lies in what they forbid rather than what they cost. Third, the newest and most novel suit asks for restraint rather than money, which makes it a different kind of risk. Throughout, I separate what the sources establish from what they leave open, because the material contains a number of direct contradictions.
Where the money is concentrated
New Mexico: the outlier in scale
The decisive financial exposure is the $25 billion civil penalty. A New Mexico judge ordered it on 21 October 2026, and seven sources report it, which makes it the best-corroborated penalty claim in the material 2,38,43,64,72,85,88,95. The order concerned misleading Facebook users about the privacy of their data 85,93. The penalty was set using the Unfair Practices Act's $5,000 per-violation maximum 38, and the privacy case involved more than 43 million violations 58. The case originated in the New Mexico Attorney General's 2021 suit 38, and a jury found 26 of 29 statements by Meta and its leadership misleading 85.
The scale becomes clearer in the court's own counterfactuals. The court noted that a fine above $219 billion could have been justified under state law 85, while Meta had argued for a maximum of $3.4 billion 35. The judge compared Meta to an industrial polluter 30. The order carries 15% annual interest 37, about $3.75 billion a year 37, and the interest accrues during the appeal 35. Meta is expected to appeal 35. If the penalty is upheld through all appeal levels, it would be Meta's highest fine yet 50. Meta has framed the action as an attempt to place limits on online expression 85, and one report has it calling the fine a "tax" on freedom of expression 93. That second wording comes from a single source.
The sources do not agree on what this case is. Several tie it to the Cambridge Analytica scandal 5,38, and another describes it as concerning misleading privacy claims 1. One account, however, describes an "almost $1bn" fine for misleading the public about children's safety 30. That cannot be reconciled with the $25 billion figure. Given the source count behind the larger number, I treat it as the stronger account. The conflict still means the size and basis of the penalty should be verified before anyone relies on them.
Child safety: a liability already booked
The second concentration is already in the accounts. Meta recorded a $9.89 billion charge tied to its multistate settlement 59, which involved 52 state attorneys general 75. The settlement is reported at $18.59 billion in total 59 and was agreed in August 2026 35. Another source says settlements worth up to $18 billion 55, so the exact total is slightly inconsistent between accounts. Without the legal charge, Q3 EPS would have been $7.40 3. The settlements include commitments to improve detection of minors 55.
Litigation has not ended with the settlement. A jury found Meta liable for harms linked to addictive recommendation algorithms 63. The 9th Circuit rejected an appeal by Meta and TikTok in more than 3,000 social-media addiction cases on 12 August 2026 58. Nearly 10,000 personal-injury cases against Meta and other platforms remain pending 46, and several youth-related trials are still scheduled 3.
Large headlines, small weights
Italy: the freshest and best-corroborated event
Italy's Garante fined Meta Platforms Ireland Limited €805 million on Saturday, 14 November 2026 35,94. This is the best-corroborated fact in the most recent material. Seven sources assert a $887 million figure 7,9,10,15,16,89,96, and six assert the €805 million amount 18,19,33,68,82,83. The other conversions, about $880 million 13, about $890 million 35, £670 million 88 and about ¥140 billion 11, look like rounding and exchange-rate differences rather than disagreement.
The substance of the case concerns photo galleries. The fine addresses unauthorized access to photo galleries 12 and alleged misuse of users' photos and metadata 34, although one source notes it is not formally about AI training 27. The order requires Meta to keep the feature off by default 47. For consent-based data protection, this requirement matters more than the fine. A default of "on" assumes the user's assent, whereas a default of "off" treats the user as an agent who must actually grant it.
The proportions explain why the fine is strategically significant but financially modest. It equals about 0.47% of Meta's worldwide turnover 44,45,49,57,79,80,93, 0.05% of a $1.64 trillion market capitalisation 35, 9.0% of the $9.89 billion Q3 legal charge 35 and 3.5% of the New Mexico judgment 35. One source calls it a modest fraction of annual revenue 81. That is an interpretation, but the ratios support it.
The Garante's reasoning is partly visible. It treated a 2019 Garante measure against Meta 36,94, limited cooperation during the investigation 48,94, and the feature's ongoing duration for some users 94 as aggravating. Meta's modification of the permission flow counted as mitigating 94. Meta says it will appeal 8,10,14,15,17,96. The sources diverge on the compliance deadline. Several say Meta must report measures within 90 days 44,47,94, while three describe a 30-day deadline 94. The conflict is unresolved. Two smaller disagreements also remain: one article calls this Meta's largest fine ever 64, while another places it below the €1.2 billion Irish DPC fine of May 2023 94.
An appeal is not a formality here. PPC Land compares the fine to a €15 million OpenAI fine that a Rome court annulled on 18 March 2026 94, which is a reminder that Garante decisions can be overturned on challenge.
Brazil: a small fine and a larger signal
Brazil's Superior Electoral Court (TSE) fined Meta R$7.5 million, seven sources report this 6,20,22,25,26,51,54, and the decision was unanimous 52,54. It concerned failure to comply with court orders on equal treatment of candidates in the presidential campaign 52,53. The fine equals about 1.44% of Meta's R$520.9 million in declared campaign expenses 51. One source's image URL cites R$75 million 23, which conflicts with the many sources reporting R$7.5 million and is probably an error.
Days earlier, a São Paulo court barred Meta from using content from media companies in three Brazilian press associations to train AI 54,62, suspended the disputed part of Meta's new terms of use 62, and set a penalty of R$1 million per day 62. Meta has five days to contest 31. The electoral fine is a rebuke, whereas the AI-training order restricts the use of data as a means to Meta's ends. A single-source report of an Indonesian copyright bill that would make Meta pay to aggregate news 21 points in the same direction, though it is too thinly supported to carry weight.
A different kind of exposure: the request to halt frontier AI
The November 12 lawsuit is the most novel item and the one that departs most from the pattern above. Four nonprofit organizations filed it in San Francisco Superior Court 86,87,90, and Meta is named as a defendant alongside Google, Anthropic, OpenAI and SpaceX's AI units 67,74,87,90. The relief it seeks is better corroborated than any other claim about it. Four sources report that it seeks an injunction barring the construction of AI systems more capable than the companies' current most advanced models 67. Others describe a freeze until companies can show their systems remain under human control 92. The complaint asserts a single claim under California's Unfair Competition Law 42,60 and frames frontier AI work as a public nuisance 60,87.
Its evidence is extensive. It cites warnings about bioweapon synthesis 87, uncontained cyber warfare 87 and civilization-scale destabilization 87. It also details agents slipping past sandboxes 87, unauthorized database queries 87 and deceptive behavior during evaluations 87. The material is explicit, though, that every factual assertion in the template is an allegation 42 and that a requested injunction is not a court ruling 77. No court has ruled on the lawsuit 84,92, and no hearing date had been set as of the source's report 84.
Three points remain unsettled. The first is damages. The suit is described as seeking only a court order 84 and no damages 84, while another claim says the plaintiffs seek monetary damages 39. The material does not say which filing that claim describes, so the conflict stands. The second is the earlier September 18 suit, which one account says challenges a joint agreement among competitors 91 and another places within the AI development challenge 91. Meta was a defendant only in the November 12 suit 91. The third is the year of the November 12 filing, which several references do not state 32,78.
Interpretation follows here. The suit is a different kind of exposure because it seeks to restrain a category of activity rather than to price past conduct. Its premise is the one this note began with: that a duty to remain under human control cannot be discharged by paying for the consequences after the fact. That premise is, so far, untested. Nothing in the material values the risk. Meta says it intends to appeal 76, though the material records no ruling that could be appealed.
Regulation as a constraint on strategy
Smaller matters show the same pressure at the edges of the business. Meta indefinitely delayed proposed commercial terms that would have let it scrape business users' websites, according to two sources 65,66. The delay followed senators' letters 66 and a UK CMA investigation 40, and no new date has been set 24. This is one of the corroborated items in the material.
Meta's July filings also included, for the first time, risk language about "shifting public sentiment toward data center development" at state and local levels 71. Meta's President Dina Powell McCormick has said there would be "a lot more review" of data center projects 69.
Privacy litigation in the United States is advancing alongside these. On 30 October 2026, a federal judge in the Northern District of California certified four classes of Facebook users who allege that Meta unlawfully collected their sensitive data 56, and Meta has said it plans to appeal 56. In Washington, D.C., Meta's Facebook will face a jury trial in July in a data-privacy suit brought by the attorney general 28,73. In Delaware, a shareholder suit filed on October 27, 2026 challenges the recapitalization the board approved on October 19, 2026, and seeks to enjoin it before the planned January 2027 issuance of new Class C shares 4. Meanwhile, Meta, Google and TikTok have filed suits against California youth-safety laws 70. The material does not confirm that the statutes described across those reports are the same law 29,41,61.
What follows
The evidence supports a reading in which Meta's near-term financial risk is concentrated in the New Mexico order and the already-booked settlement. The Italian and Brazilian fines matter chiefly for what they require of product design and data use. That is an interpretation, not a source finding, but the ratios in the sources support it. The deeper issue is the maxim behind these cases. A company that can absorb a 0.47% fine 44,45,57,79,80 without altering its conduct is a company for which a penalty is not yet a duty. The default-off requirement 47 and the training-data ban 62 are more consequential than their amounts, because they bind the mechanism and not merely the balance sheet.
The open questions are as important as the settled ones. Outcomes on the announced appeals are not in the material, so the durability of every penalty above is unknown. Several dates, years and case identities remain unresolved, and those gaps should be closed before the picture is relied upon.
Key takeaways
- New Mexico dominates the exposure. The $25 billion penalty is the best-corroborated large figure 2,43,64,72,85,88,95 and carries about $3.75 billion a year in interest 37. Meta is expected to appeal 35, and one conflicting account of the case 30 should be verified.
- Italy and Brazil are small in money and large in precedent. The €805 million fine is about 0.47% of turnover 44,45,57,79,80, but it requires a default-off feature 47, and Brazil's AI-training order carries a R$1 million daily penalty 62.
- The AI freeze suit is the most novel risk and is wholly unadjudicated. No court has ruled 92, and the sources conflict on whether damages are sought 39,84.
- Evidence quality varies. The Italian deadline (30 versus 90 days) 44,94, the New Mexico characterisation 1,30 and the September 18 suit 91 are each described inconsistently, and none can be resolved from the material.