Report on Foundation Model Impacts Released

Why It Matters

Transparent reporting of foundation‑model impacts is critical for managing societal risks and informing regulators, investors, and the public. The findings signal that industry‑wide standards are still needed to ensure responsible AI deployment.

Key Takeaways

• 13 major AI firms evaluated on four governance practices
• Only a few providers actively share post‑deployment impact data
• Usage metrics improve, but incident reporting remains sparse
• Report highlights need for standardized, comprehensive impact disclosures

Pulse Analysis

Foundation models—large‑scale AI systems powering chatbots, image generators, and code assistants—have moved from research labs to commercial products at unprecedented speed. As these models embed themselves in critical workflows, questions about their societal effects, from labor displacement to misinformation, have intensified. Partnership on AI, a multi‑stakeholder nonprofit, has positioned itself as a watchdog by publishing annual transparency reports that benchmark how developers disclose real‑world outcomes. The 2026 edition expands the assessment framework to 19 granular processes, reflecting a maturing methodology that blends academic rigor with industry insight.

The report’s authors, Jacob Pratt and Albert Tanjaya, sifted through more than 150 papers, articles, and public disclosures to gauge progress across four core practices: sharing usage data, publishing impact‑indicator research, reporting incidents and policy breaches, and aggregating user feedback. While usage statistics and labor‑market impact analyses have become more visible, the study reveals a stark deficiency in reporting serious harms, incident frequencies, and broader societal repercussions. Only a minority of the 13 evaluated providers—among them OpenAI, Google, and Microsoft—demonstrate systematic, transparent reporting, leaving regulators and civil society with an incomplete risk picture.

For businesses and policymakers, the report underscores an urgent need for standardized impact reporting regimes. Consistent, comparable data would enable investors to assess AI‑related risks, help regulators craft evidence‑based policies, and empower users to make informed choices. As the AI ecosystem continues to scale, the pressure will mount on lagging firms to adopt the best practices highlighted in the report, turning transparency from a voluntary badge into a market imperative.