Unicode 18.0 arrives with nine emoji and three new currencies

Unicode 18.0 dropped on 16 September, and the headlines write themselves: nine new emoji, three new currency symbols, 13,007 new characters. The part worth paying attention to, though, is a quiet rewording of the conformance rules covering variation selectors — the invisible code points that tell a system how to render the character before them. According to the Unicode Consortium's own announcement, the updated language exists partly because research has shown that long sequences of invisible variation selectors can be used to attack modern AI applications.

What is actually in Unicode 18.0

The release adds 13,007 characters for a total of 172,808, and for once the biggest addition is not an emoji. The largest new set is Seal, the script the code charts label “Small Seal”: 11,328 ideographic characters that date back to China's Qin Dynasty, around 200 BCE. Jurchen, used in northeastern China during the Jin Dynasty, is the second new ideographic script, and Proto-Cuneiform arrives with its archaic numerals only — additional signs are promised for a future version. Seven new blocks were defined to hold it all, from Bengali Supplement and Archaic Cuneiform Numerals through to Musical Symbols Supplement, Miscellaneous Symbols and Arrows Extended, and a block for Seal itself.

There are no new algorithms this time. Instead the consortium has published a new Unicode Standard Annex, UAX #60, covering the data files for East Asian scripts, and added JurchenSources.txt and SealSources.txt alongside the existing Nushu and Tangut source files. Five technical standards that version in step with the standard itself have been refreshed at the same time: the collation algorithm, the security mechanisms, IDNA compatibility processing for non-ASCII domains, the emoji specification and the link detection standard. The collation and IDNA updates are the ones that touch European software most directly, since they govern how text sorts and how internationalised domain names resolve.

Three currency symbols that had to wait for a standards meeting

Three of the new characters are currency signs: U+20C2 RUFIYAA SIGN, U+20C3 UAE DIRHAM SIGN and U+20C4 OMANI RIAL SIGN. All three were authorised for public use by their respective monetary authorities more than a year before this release, and, as the consortium notes, practical use was held back by the absence of a standard encoding. That is the unglamorous reality of a universal character set: a country can sign off on a syllable, and its keyboards still cannot type it. Vendors are now free to implement all three, which in practice means waiting for operating systems, fonts and phones to catch up.

Nine emoji, including a monarch butterfly and a pickle

The emoji set grew by nine characters, plus ten skin-tone variants, for nineteen new sequences in total. The new arrivals are a cracking face, leftwards and rightwards thumb signs, a monarch butterfly, a pickle, a lighthouse, a meteor, an eraser and a net with handle. It is a functional batch rather than a novelty one — the eraser, the pickle and the two thumb signs all fill obvious conversational gaps — and it keeps the emoji repertoire's long-running drift towards the practical rather than the cute.

The invisible characters that can be turned on AI systems

Variation selectors have been in Unicode for years, and the overwhelming majority of honest uses are typographic. Version 18.0 tightens the definitions and requirements around them, and adds recommendations so that non-conformant uses are made visible rather than silently ignored. The consortium ties this directly to security research showing that strings of invisible variation selectors can be aimed at modern AI applications. For anyone shipping a chatbot, a document pipeline or a moderation tool, that is the paragraph of the release notes to actually read. A character you cannot see is a remarkably good place to hide an instruction.

Our opinion

The Unicode Consortium moves at a pace that makes a government consultation look sprightly, and this release is a good illustration of why that is mostly a feature. Encoding 11,328 characters from a script that went out of everyday use two millennia ago is not a commercial proposition, and the payoff is that a museum, a philologist or a typeface designer in 2050 will be able to type them without asking anyone's permission. Nobody will write a thinkpiece about the collation and IDNA updates either, yet they are the reason a Cyrillic address and a Greek-language database sort the way their readers expect. The variation selector work is the sharpest example of the standard earning its keep: the consortium spotted that an old, boring, invisible feature had become an attack surface for AI systems and clarified the rules rather than leaving it to vendors to improvise. The fair criticism is speed at the other end of the pipeline. Three currency symbols sat authorised and unusable for more than a year because no one had encoded them, and they will now sit encoded and unusable for another year or two while software catches up. That lag is the price of a standard nobody is allowed to own — and, on balance, still cheaper than the alternative.