On May 27, 2026, Italy's communications authority AGCOM adopted Delibera 127/26/CONS, establishing a permanent technical table on artificial intelligence, copyright, and information pluralism. The table invites search engines, digital platforms, news publishers, and trade associations to negotiate text-and-data-mining (TDM) opt-outs, publisher compensation for AI's use of editorial content, and shared standards against "algorithmic hallucinations" — AI-generated summaries that misstate or fabricate facts drawn from news content.
The timing is not incidental. One month earlier, on April 29, 2026, AGCOM's Council voted to refer Google Ireland Ltd.'s AI Overviews and AI Mode to the European Commission under Articles 27, 34, and 35 of the Digital Services Act, invoking the DSA's Article 65 escalation mechanism for very large online platforms (AGCOM press release, April 30, 2026). The complaint, filed by Italy's publisher federation FIEG, alleges AI Overviews suppress traffic to news sites while surfacing unverified, occasionally fabricated summaries in their place. AGCOM created the technical table at the same Council meeting — a negotiated track running in parallel with a formal regulatory escalation.
The case for the table
The steelman here is real. Generative search overviews are a structurally different threat to publishers than the display-advertising erosion of the 2010s: when an AI summary answers the query directly, the click that used to fund the reporting never happens. FIEG's complaint reflects a genuine collective-action problem — no single Italian outlet has leverage to negotiate TDM terms with Google alone, and a voluntary process convened by a regulator can produce faster, more tailored outcomes than litigation or a multi-year EU rulemaking. AGCOM's own framing, as reported by Newslinet, is that generative AI no longer merely distributes content but autonomously synthesizes it into answers users treat as complete — a fair description of what AI Overviews actually does, and a legitimate basis for regulatory attention on pluralism grounds, not just copyright.
AGCOM also isn't inventing a rulemaking gap. Italy already has statutory footing: Law 132/2025, in force since October 10, 2025, added Article 70-septies to Italian copyright law, extending the EU's TDM opt-out regime to generative AI systems and criminalizing TDM that ignores a rights holder's reservation. The technical table's job is to operationalize that opt-out — building the machine-readable signals, licensing templates, and dispute channels the statute assumes but doesn't specify.
Where the approach runs into trouble
The design still carries the two familiar risks of regulator-convened bargaining. First, sequencing: AGCOM opened a DSA Article 65 referral against Google one day before creating a table Google is being asked to join voluntarily. Whatever the participants' good faith, a company under active EU Commission scrutiny negotiates differently than one meeting a regulator on equal footing — the leverage asymmetry is built into the calendar, not the substance.
Second, scope creep from copyright into content adjudication. TDM opt-outs and compensation are contract and property-rights questions AGCOM can reasonably help broker. "Algorithmic hallucinations," by contrast, is a factual-accuracy standard, and a state communications regulator sitting at the table where that standard gets defined edges toward AGCOM adjudicating what counts as acceptable AI output about the news — a role Italy's own media pluralism norms are supposed to keep at arm's length from government. AGCOM board member Elisa Giomi made a version of this point publicly, reportedly dissenting from the EU referral on the grounds that protecting publishers' commercial interests isn't AGCOM's mandate — a notable internal crack in an otherwise unanimous-seeming initiative.
The proportionate path
A table that produces machine-readable opt-out standards and a transparent compensation formula — the kind of interoperable infrastructure Italy's Law 132/2025 already anticipates — would be a genuine public good, cheaper and faster than either litigation or a new EU-level rule. A table that becomes an alternative venue for defining what AI is allowed to say about the news, running alongside an active DSA enforcement referral against one of its own invited participants, blurs into something closer to supervised content policy. AGCOM has given itself a wide mandate and a full calendar; whether this becomes a template for interoperable AI-publisher licensing or a slow-motion pluralism inquiry with Google in the room depends entirely on which half of that mandate the table actually uses.