1% generated by tables/make_tables.py from ./results/graph/category_cooccurrence_yearly.csv, ./results/graph/category_counts_yearly.csv2\begin{longtable}{lrrrrrrrrrrrrrr}3\caption{Co-classification of security-and-privacy works with the LLM, AI-safety and AI-agents categories, by publication year: number of works carrying both categories and share of the security category's works.}\label{tab:security_llm_cooccurrence}\\4\toprule5\textbf{Year} & \textbf{Cyber. x LLM} & \textbf{Cyber. x LLM (\%)} & \textbf{Cyber. x AI safety} & \textbf{Cyber. x AI safety (\%)} & \textbf{Cyber. x AI agents} & \textbf{Cyber. x AI agents (\%)} & \textbf{Crypto. x LLM} & \textbf{Crypto. x LLM (\%)} & \textbf{Blockch. x LLM} & \textbf{Blockch. x LLM (\%)} & \textbf{Privacy x LLM} & \textbf{Privacy x LLM (\%)} & \textbf{Fed. learn. x LLM} & \textbf{Fed. learn. x LLM (\%)} \\6\midrule7\endfirsthead8\toprule9\textbf{Year} & \textbf{Cyber. x LLM} & \textbf{Cyber. x LLM (\%)} & \textbf{Cyber. x AI safety} & \textbf{Cyber. x AI safety (\%)} & \textbf{Cyber. x AI agents} & \textbf{Cyber. x AI agents (\%)} & \textbf{Crypto. x LLM} & \textbf{Crypto. x LLM (\%)} & \textbf{Blockch. x LLM} & \textbf{Blockch. x LLM (\%)} & \textbf{Privacy x LLM} & \textbf{Privacy x LLM (\%)} & \textbf{Fed. learn. x LLM} & \textbf{Fed. learn. x LLM (\%)} \\10\midrule11\endhead122018 & 536 & 0.8 & 929 & 1.4 & 571 & 0.8 & 66 & 0.2 & 263 & 2.4 & 58 & 0.5 & 13 & 1.2 \\132019 & 735 & 1.0 & 1{,}726 & 2.4 & 621 & 0.8 & 69 & 0.2 & 272 & 1.8 & 85 & 0.7 & 15 & 1.0 \\142020 & 985 & 1.3 & 2{,}648 & 3.5 & 639 & 0.8 & 108 & 0.3 & 281 & 1.7 & 130 & 0.8 & 36 & 1.3 \\152021 & 1{,}231 & 1.6 & 3{,}438 & 4.4 & 645 & 0.8 & 136 & 0.4 & 283 & 1.6 & 233 & 1.3 & 66 & 1.6 \\162022 & 1{,}337 & 1.7 & 4{,}056 & 5.2 & 596 & 0.8 & 198 & 0.6 & 275 & 1.4 & 278 & 1.4 & 91 & 1.5 \\172023 & 2{,}304 & 2.6 & 5{,}288 & 5.9 & 807 & 0.9 & 403 & 1.1 & 357 & 1.5 & 717 & 2.8 & 226 & 2.7 \\182024 & 4{,}977 & 4.6 & 7{,}850 & 7.3 & 1{,}233 & 1.1 & 839 & 2.0 & 508 & 1.9 & 2{,}022 & 5.8 & 601 & 5.0 \\192025 & 10{,}549 & 7.6 & 12{,}933 & 9.3 & 3{,}207 & 2.3 & 1{,}745 & 3.4 & 1{,}033 & 3.1 & 4{,}584 & 9.3 & 1{,}374 & 7.6 \\202026 & 14{,}410 & 12.0 & 14{,}029 & 11.6 & 6{,}329 & 5.3 & 2{,}423 & 5.5 & 1{,}067 & 4.3 & 5{,}665 & 13.6 & 1{,}990 & 11.7 \\21\bottomrule22\multicolumn{15}{p{0.95\linewidth}}{\footnotesize A work is co-classified when the multi-label classifier assigns both categories (Part IV). Shares are relative to all eligible works carrying the security category in the year (multi-label count). Non-zero counts before 2022 reflect the breadth of the keyword patterns of the AI categories (for example `language model' matches pre-transformer work), not the use of LLMs; the table should be read for its trend, not its level. The 2026 row is incomplete.}\\23\end{longtable}24