01 — 2025 SURVEY

2025 survey

The 2025 State of Production ML report analyses 135 practitioner responses submitted from 31 August to 1 December 2025. It compares matching questions with the 2024 baseline and reports changes in production challenges, tools, infrastructure and organisational practice.

The respondents span roles, countries, industries and organisation sizes across the production machine-learning ecosystem. This is a self-selected practitioner sample, so its findings describe respondents rather than population-wide prevalence.

Cite this survey as

The Institute for Ethical AI Alignment & Safety (2025), The State of Production ML in 2025, https://ethical.institute/reports/state-of-ml-2025/.

02 — 2024 SURVEY

2024 survey

The 2024 State of Production ML report analyses 177 practitioner responses submitted from 14 September 2024 to 15 January 2025. It is the baseline for the year-on-year comparisons in the 2025 edition.

The survey covers the same production context, tooling, organisation and respondent-profile areas used by the later edition. Its respondent mix is shown within the report so each result can be read against the sample that produced it.

Cite this survey as

The Institute for Ethical AI Alignment & Safety (2024), The State of Production ML in 2024, https://ethical.institute/reports/state-of-ml-2024/.

03 — METHODOLOGY

Methodology

Both reports aggregate the committed survey CSV at build time. Each percentage uses the number of nonblank responses to that question, so the denominator varies across charts. For questions allowing multiple selections, percentages are the share of respondents selecting each answer and can sum to more than 100%.

The reports publish aggregates rather than raw response tables because the source rows include free-text and contact fields. Field dates above are the earliest and latest timestamps in each committed dataset.