Based off PimDeWitte's, this improves performance by over a magnitude and eliminates all garbage (allocations).

BadWords.java

import it.unimi.dsi.fastutil.longs.Long2ObjectMap;
import it.unimi.dsi.fastutil.longs.Long2ObjectOpenHashMap;
import net.openhft.hashing.LongHashFunction;

import java.io.BufferedReader;
import java.io.IOException;
import java.io.InputStreamReader;
import java.net.URL;

/**
 * Originally created by Pim De Witte.
 * 
 * Performance drastically improved by over an order of magnitude by Thomas G. P. Nappo (Jire).
 * Garbage production has been eliminated as well.
 */
public class BadWords {
	
	static Long2ObjectMap<String[]> words = new Long2ObjectOpenHashMap<>();
	static int largestWordLength = 0;
	
	public static void flag(String word) {
		String[] ignore_in_combination_with_words = new String[]{};
		if (word.length() > largestWordLength) {
			largestWordLength = word.length();
		}
		words.put(LongHashFunction.xx().hashChars(word.replaceAll(" ", "")), ignore_in_combination_with_words);
	}
	
	public static void loadConfigs() {
		try {
			BufferedReader reader = new BufferedReader(new InputStreamReader(new URL("https://docs.google.com/spreadsheets/d/1hIEi2YG3ydav1E06Bzf2mQbGZ12kh2fe4ISgLg_UBuM/export?format=csv").openConnection().getInputStream()));
			String line = "";
			int counter = 0;
			while((line = reader.readLine()) != null) {
				counter++;
				String[] content = null;
				try {
					content = line.split(",");
					if(content.length == 0) {
						continue;
					}
					String word = content[0];
					String[] ignore_in_combination_with_words = new String[]{};
					if(content.length > 1) {
						ignore_in_combination_with_words = content[1].split("_");
					}
					
					if(word.length() > largestWordLength) {
						largestWordLength = word.length();
					}
					words.put(LongHashFunction.xx().hashChars(word.replace(" ", "")), ignore_in_combination_with_words);
					
				} catch(Exception e) {
					e.printStackTrace();
				}
				
			}
			System.out.println("Loaded " + counter + " words to filter out");
		} catch (IOException e) {
			e.printStackTrace();
		}
	}
	
	private static final char[][] convert = {
			{'o', '0'},
			{'i', '1'},
			{'l', '1'},
			{'t', '+'},
			{'e', '3'},
			{'i', '!'},
			{'l', '!'},
			{'s', '$'},
			{'a', '&'},
			{'a', '@'},
			{'c', '('},
			{'d', ')'},
			{'d', '0'},
			{'g', '6'},
			{'t', '7'},
			{'g', '9'},
			{'s', '5'},
			{'a', '4'}
	};
	
	private static final ThreadLocal<StringBuilder> sb = ThreadLocal.withInitial(StringBuilder::new); // make this regular if you don't need thread safety.
	
	/**
	 * Iterates over a String input and checks whether a cuss word was found in a list, then checks if the word should be ignored (e.g. bass contains the word *ss).
	 *
	 * @param input
	 * @return
	 */
	public static boolean badWordsFound(String input) {
		if (input == null) {
			return false;
		}
		
		StringBuilder sb = BadWords.sb.get();
		sb.setLength(0);
		
		removeLeetspeak:
		for (int i = 0; i < input.length(); i++) {
			char c = input.charAt(i);
			if (Character.isLetter(c)) {
				sb.append(Character.toLowerCase(c));
			} else {
				for (char[] conversion : convert) {
					if (c == conversion[1]) {
						sb.append(conversion[0]);
						continue removeLeetspeak;
					}
				}
			}
		}
		
		// iterate over each letter in the word
		for (int start = 0; start < sb.length(); start++) {
			// from each letter, keep going to find bad words until either the end of the sentence is reached, or the max word length is reached.
			for (int offset = 1; offset < (sb.length() + 1 - start) && offset < largestWordLength; offset++) {
				long hash = LongHashFunction.xx().hashChars(sb, start, offset);
				if (words.containsKey(hash)) {
					// for example, if you want to say the word bass, that should be possible.
					String[] ignoreCheck = words.get(hash);
					boolean ignore = false;
					for (int s = 0; s < ignoreCheck.length; s++) {
						if (indexOf(sb, ignoreCheck[s]) >= 0) {
							ignore = true;
							break;
						}
					}
					if (!ignore) {
						return true;
					}
				}
			}
		}
		
		return false;
	}
	
	private static int indexOf(CharSequence source, CharSequence target) {
		int sourceCount = source.length();
		int targetCount = target.length();
		int sourceOffset = 0;
		int targetOffset = 0;
		
		if (0 >= sourceCount) {
			return (targetCount == 0 ? sourceCount : -1);
		}
		if (targetCount == 0) {
			return 0;
		}
		
		char first = target.charAt(targetOffset);
		int max = sourceOffset + (sourceCount - targetCount);
		
		for (int i = sourceOffset; i <= max; i++) {
			/* Look for first character. */
			if (source.charAt(i) != first) {
				while (++i <= max && source.charAt(i) != first);
			}
			
			/* Found first character, now look at the rest of v2 */
			if (i <= max) {
				int j = i + 1;
				int end = j + targetCount - 1;
				for (int k = targetOffset + 1; j < end && source.charAt(j)
						== target.charAt(k); j++, k++);
				
				if (j == end) {
					/* Found whole string. */
					return i - sourceOffset;
				}
			}
		}
		return -1;
	}
	
}

I know you are about full memory optimization, but is there any way to return the swear word for a reference in an automatic report? This is because I get a ton of false positives in messages and its easier to figure out what the false positive is if I have the word available.

ie
if (!ignore) { return wordToCheck; }
and at the end if nothing is found return null