ESG Ratings vs Court Records: Why They Disagree (2026)

Why Do ESG Ratings Disagree With Court Records?

Meta carries an MSCI ESG rating of AA -- the second-highest tier. Courts and regulators have documented billions in penalties against the company across three years, spanning unlawful data transfers, illegal facial recognition use, and forced advertising consent.

Microsoft holds MSCI's highest rating: AAA. An estimated 11.9% of its annual revenue comes from defence contracts, including a $22 billion HoloLens augmented reality contract with the U.S. Army.

These are not edge cases. They are the norm. A 2022 study by Berg, Koelbel, and Rigobon at MIT Sloan found that the correlation between six major ESG rating agencies was just 0.54. For context, credit rating agencies agree about 99% of the time. ESG raters, using nominally similar inputs, produce outputs that diverge nearly as often as they align.

We took six of the most widely held public companies, compared what ESG raters say about them to what courts, regulators, and journalists have documented, and measured the gap. The structural difference explains why ESG labels and corporate conduct often tell different stories.

For a look at what this means inside actual funds, see our analysis of 10 popular ESG ETFs. We also examined what's really inside "sustainable" funds.

What ESG Ratings Measure vs. What Courts Document

The divergence is not random. It traces back to input selection -- what data each system treats as primary evidence.

ESG ratings draw primarily on:

• Corporate sustainability reports and voluntary disclosures
• Company-completed questionnaires and surveys
• Policy documents, governance structures, and stated commitments
• News monitoring services with variable incident coverage

Independent scoring (Mashinii's approach) draws on:

• Court filings and settled litigation
• Regulatory penalties and enforcement actions
• Investigative journalism from established outlets
• NGO field reports and documented findings

The first system measures what companies say about themselves. The second measures what external authorities have documented about their conduct. Both capture real information. They are not interchangeable.

This creates three measurable problems:

1. Self-reported data is inherently selective. A company can highlight its renewable energy purchases in a sustainability report while not disclosing a $14 million air pollution penalty or a $725.5 million benzene exposure judgment. ESG ratings built on these disclosures inherit whatever the company chose to emphasise.

2. Different agencies weight different factors differently. MSCI might reward a company for publishing a climate policy. Sustainalytics might penalise the same company for rising carbon emissions. Both capture something real, but they disagree on what matters more.

3. Enforcement data is lagging and inconsistently incorporated. When rating agencies do incorporate negative events, they often rely on news monitoring with variable latency and coverage. A headline-making billion-dollar fine may be captured quickly. A pattern of labour violations across 15 facilities may take years to register, if it registers at all.

The Correlation Problem: 0.54 Between Rating Providers

The Berg, Koelbel, and Rigobon study remains the most cited analysis of ESG rating divergence. Their finding: across six major ESG rating agencies, the average pairwise correlation was 0.54. The researchers attributed 53% of the divergence to "measurement" -- the same category assessed using different indicators -- and 33% to "scope" -- which categories are included at all.

To illustrate: MSCI rates Meta as AA ("Leader"). Sustainalytics rates Meta's ESG risk as "High." Both are commercially available ratings used by fund managers to construct ESG portfolios. An investor consulting one provider gets a fundamentally different signal than an investor consulting the other.

If a portfolio construction decision depends on which provider you happen to use, the label is not conveying reliable information about the underlying company.

5 Companies Where ESG Ratings and Independent Data Diverge Most

We selected five companies that are widely held, heavily rated by ESG providers, and the subject of well-documented regulatory activity. For each, the divergence between ESG consensus and independently documented conduct is measurable.

Mashinii scores range from -100 to +100. Negative scores indicate documented evidence of harm outweighs evidence of positive action. ESG ratings sourced from publicly available provider summaries.

What the Divergences Show, Company by Company

Tesla: The Product Is Not the Company

Tesla makes electric vehicles that reduce air pollution. Zero-emission vehicle adoption is linked to a 3.2% drop in asthma-related emergency visits for every additional 20 ZEVs per 1,000 people. Some ESG raters reward this accordingly.

But in 2022, S&P removed Tesla from its ESG index. The independent record shows why: $7.4 million in ethics-related fines, Foreign Corrupt Practices Act penalties totalling $14.3 million, over 112 air pollution permit violations at its Fremont facility in five years, and at least 15 substantiated labour-law violations in 2023 and 2024.

ESG ratings that emphasise the product category may not capture the operating record. Independent evidence scores the company, not the concept.

View Tesla's full score breakdown

Meta: AA From One Provider, "High Risk" From Another

Meta carries MSCI's second-highest ESG rating. Sustainalytics classifies it as "High Risk." The same company, rated by two of the largest ESG providers, receives conflicting signals.

The independently documented record includes a record EUR 1.2 billion GDPR fine for unlawful data transfers, a $1.4 billion facial recognition settlement in Illinois, and EUR 91 million for storing passwords in plain text. Thirty-three U.S. states allege the company's platforms are addictive and harmful to youth mental health. Meta's Oversight Board criticised the company for leaving up anti-Muslim and anti-migrant content during the 2025 UK riots.

The gap between an AA ESG rating and the enforcement record reflects what happens when the rating primarily reflects corporate disclosure rather than regulatory outcomes.

View Meta's full score breakdown

Microsoft: AAA Rating, Undisclosed Defence Revenue

Microsoft holds MSCI's highest ESG rating. On several dimensions, the data supports strong scores -- verified pay equity globally and +50 on Respect for Cultures & Communities.

But ESG frameworks largely do not surface Microsoft's defence exposure. An estimated 11.9% of annual revenue comes from defence contracts, including the $22 billion HoloLens contract. OpenAI revised its terms to allow military applications. Microsoft disbanded its AI ethics and society team in March 2023.

For investors who track military involvement, an AAA ESG rating does not convey the extent of defence revenue. The rating measures what was disclosed. Independent data measures what courts and procurement records document.

View Microsoft's full score breakdown

Amazon: "Above Average" ESG, Documented Worker Conditions

Amazon carries an MSCI rating of A, placing it above average. The company will pay $2.5 billion to settle a U.S. lawsuit over deceptive Prime subscription practices. France fined it EUR 32 million for intrusive employee monitoring. The NLRB found Amazon illegally enforced its Unpaid Time Off policy in February 2025. Three substantiated forced labour findings since 2022 include connections to Uyghur forced labour in Xinjiang.

Amazon scores +50 on Respect for Cultures & Communities, having invested $2.2 billion to create or preserve over 21,000 affordable homes. The independent record is not uniformly negative -- but the "Above Average" ESG label does not convey the documented labour record.

View Amazon's full score breakdown

Alphabet: Deleted Its AI Ethics Commitment After Receiving ESG Ratings

Alphabet scores -80 on No War, No Weapons. In February 2026, Google rewrote its AI principles, removing its pledge not to use AI for weapons or surveillance. This followed the $1.2 billion Project Nimbus contract and a share of the $9 billion Joint Warfighting Cloud contract with the Pentagon. When employees protested, Google terminated 28 of them.

ESG ratings that were calibrated when Google had an AI weapons pledge have not necessarily been revised to reflect the pledge's removal. The lag between a company's policy change and its ESG rating update is a structural feature of disclosure-based systems.

View Alphabet's full score breakdown

What This Means for Investors Relying on ESG Labels

Across all five companies, the data illustrates a consistent structural difference: ESG ratings primarily measure policy and disclosure. Independent data primarily measures conduct and consequence.

ESG ratings are not without value. They capture genuine information about governance structures and stated environmental commitments. But they are incomplete for investors who want to understand documented corporate behaviour -- especially on dimensions like labour practices, military involvement, data privacy, and ethical sourcing where self-reported data diverges most from enforcement records.

Three practical observations:

No single ESG score captures the full picture. If two major rating agencies classify the same company as "Leader" and "High Risk" respectively, neither label alone is sufficient for a values-aligned investment decision.

Independent sources provide structurally different information. Court filings, regulatory actions, and investigative journalism document what companies do. Corporate self-reporting describes what companies say they do. Both are useful. They are not interchangeable.

Specific dimensions matter more than averages. A company can score well on climate and poorly on labour rights. An investor who cares about worker welfare needs data that measures worker welfare -- not a composite score that averages it with other dimensions.

Scoring methodology

Mashinii scores companies across 11 ethical dimensions using independently sourced evidence -- court records, regulatory actions, investigative reports, and NGO findings. Every score is backed by cited sources.

Learn more about our methodology

See the Full Picture

Search any company or audit your portfolio to compare your holdings against independent data from courts, regulators, and journalists.

Advisors building client portfolios under the FCA's Anti-Greenwashing Rule can use Mashinii's court-filing-backed scores in place of corporate self-disclosures. Explore the advisor solution.

Audit My Portfolio | Search Any Company | View Rankings

---

Mashinii provides integrity data for informational purposes only. Nothing on this page constitutes financial, legal, or investment advice.