Security and safety was the largest responsible AI research topic in 2025, with 641 papers, up 23%

What the chart shows

Security and safety was the largest responsible AI subtopic at six major AI conferences in 2025, with 641 accepted papers, up 23% from 521 in 2024. Fairness and bias followed with 462 (up 13%), then transparency and explainability with 405 (up 14%).

Privacy and data governance was the smallest subtopic, at 248 papers, but grew fastest, up 33% from 186. All four subtopics are larger than in 2019; security and safety has added the most papers since then, up from 162.

Key numbers

How to read it

• A paper can belong to more than one subtopic, so the subtopics add up to more than the number of papers (1,756 against 1,521 in 2025; related chart).
• In 2022 the data file lists 7 papers under ‘Privacy’ separately; we add them to privacy and data governance, as the report’s totals do.
• The four subtopics are not exhaustive.