Apple is negotiating with news publishers over content-licensing deals reportedly worth millions of dollars, aimed at improving the AI-powered version of Siri that Apple has promised to ship in 2026. According to AIN.ua, the talks are ongoing, and neither the publishers involved nor the exact terms have been disclosed.
The move comes after nearly two years of public struggle with Siri's AI rebuild. Apple unveiled a more conversational, context-aware Siri at WWDC 2024 as part of Apple Intelligence, then delayed it repeatedly, eventually admitting the feature needed more time and reshuffling the team responsible for it. A version overhauled around large language models is now expected later in 2026.
For a company that built its AI strategy on-device and on data it already controls, paying publishers directly for content is a notable departure — and a sign of how hard it is to make a consumer assistant reliable without outside data.
Why Siri needs someone else's content
A voice assistant that answers questions about news, sports scores, or "what happened today" cannot rely purely on a model's training data, which goes stale the moment training ends. To stay current, systems like this typically retrieve fresh text from external sources and feed it into the model at query time — a pattern generally called retrieval-augmented generation, or RAG.
- Scraping the open web is cheap but legally risky and increasingly blocked by publishers themselves.
- Licensing content directly gives Apple clean rights to display snippets, summaries, or full answers sourced from that content inside Siri.
- It also gives Apple a contractual answer if a publisher later claims its work was used without permission — a problem several AI companies are still fighting in court.
Apple is late to a deal-making trend
Paying publishers for AI use rights is not new — it is the industry's default move. OpenAI has signed licensing agreements with News Corp, Axel Springer, Condé Nast, and others. Perplexity has struck revenue-share deals with publishers whose content shows up in its answers. Microsoft licenses content for Copilot. Meta reached its own agreement with the Associated Press.
What is new is Apple joining that list. The company has generally avoided paying broadly for training data, leaning instead on its device install base, first-party user data, and licensing deals scoped narrowly (music, video, and now, apparently, news). If confirmed, this deal would put Siri in the same publisher-licensing category as ChatGPT and Perplexity, rather than treating news content as freely available web material.
What this means for teams building AI products
The practical lesson for anyone shipping an AI assistant with real-time knowledge isn't about Apple specifically — it's about budget lines.
- Content licensing is now a line item, not an edge case. If your product answers questions grounded in news, finance, or specialized publisher content, factor licensing costs into unit economics the same way you factor in inference costs.
- Grounding beats bigger models for freshness. A well-licensed retrieval pipeline over current content will usually outperform a larger model trained on stale data for time-sensitive queries.
- Clean rights reduce legal exposure. Direct licensing is slower and more expensive than scraping, but it removes the single biggest legal liability facing consumer AI products today.
AiiN's takeaway
Apple's Siri relaunch has already cost the company a public credibility hit; a second stumble — this time over content rights rather than delayed features — would be harder to explain away. In our estimation, Apple is likely trying to lock in licensing agreements before the new Siri ships, rather than risk the same copyright fights that have already dragged OpenAI and Microsoft into court. Whether "millions of dollars" is enough to satisfy publishers who have negotiated far larger sums with other AI companies is the open question the coming months should answer.