Systemic Altruism
← Back to the cohort
Justice & public goodsPreliminary editorial review

AI Forensics

AI Forensics investigates opaque platform algorithms and shares auditing tools, datasets and methods with public-interest researchers. The enabling hypothesis is that common evidence infrastructure makes independent scrutiny easier.

Visit the organization ↗

Systemic Altruism practice potential: 64–74/100
90% of weighted axes reviewed. A mechanism hypothesis, not proven effectiveness or a donation recommendation.

Who this work enables

Tools & support for charities: Shared audit methods, evidence-collection infrastructure and datasets can help public-interest researchers and accountability organizations scrutinize platforms.

Humanity’s Checklist source match

Gatekeepers · Coordination — editorial, partial alignment.

Independent scrutiny can challenge platform information gatekeeping, while shared methods coordinate researchers. This is not verified enforcement, unrestricted data access or adoption of the checklist.

Explicit checklist adoption: not verified. A mechanism match is not an affiliation, endorsement, evidence of effectiveness or a scoring bonus. Read the practice.

Focus classification: 2026-10-27. Based on the organization’s dated official sources below; earlier axis-review dates are unchanged.

Compare alternatives for a funder →

The proposed pathway

Build independent evidence and reusable methods for scrutinizing powerful platforms

Six dimensions, with rationale

Structural depth

3 / 4

Independent audits target the informational power asymmetry between platforms and people affected by them, rather than providing only individual remedies.

Institutional reach

3 / 4

Methods and evidence can inform researchers, journalists and policymakers across platforms; visibility does not prove that institutions act on the findings.

Durability

2 / 4

Published methods and evidence can remain useful, but changing interfaces and platform countermeasures require ongoing access and technical work.

Scalability

3 / 4

Reusable collection and analysis methods can lower duplicate audit effort, while platform-specific methods, access restrictions and sensitive data constrain replication.

Funding additionality

Unreviewed

Unknown: developing shared audit infrastructure does not reveal the current marginal project pipeline, funding shortfall or what additional funding would uniquely enable.

Community agency

Unreviewed

Unknown: investigating harms to marginalized groups does not establish their binding authority over audit priorities, publication choices or organizational governance.

Evidence and selection

Primary-source programme and identity documentation; no independent impact validation in this review

Illustrative inclusion to examine this mechanism. Inclusion is not an external recommendation or a verified comparative ranking.

A testable next question

Can independent nonprofit auditors reproduce a defined AI Forensics audit using the released methods at lower cost than building their own pipeline, and do resulting findings lead to verifiable changes rather than publicity alone?

Review confidence: Low–medium for the documented mechanism; low for causal effectiveness and donation comparisons. The current AI Forensics association and shared algorithm-auditing work, not every historical Tracking Exposed repository, a regulator’s actions or all downstream investigations.

What remains uncertain

Scoped to the French association Artificial Intelligence Forensics, SIREN 902454313 / RNA W922018815, and the current AI Forensics programme founded in 2021. The official register reports an active declared association; that is not charitable tax advice or an impact review. Historical Tracking Exposed work is not treated as a separate current charity or a checked current software asset. No named repository license was resolved in this review, so the software-tooling tag does not carry an open-source badge. Platform access, data rights, reproducibility and regulatory causal claims remain unaudited; the image-only 2025 annual report was not used as readable legal evidence.

Sources to inspect

Propose a sourced correction →

Axis review: 2026-10-27 · weighting method v0.2 (2026-10-26) · Method, conflicts and coverage