Google, Microsoft, Meta and Anthropic harden AI-copyright risk language in October SEC filings

What happened

In SEC filings in October 2026, Alphabet, Microsoft, Meta and Anthropic all described the copyright suits over their AI training data in firmer terms than before, according to a 29 October 2026 PayAuthors summary of the filings. Alphabet's quarterly report filed 29 October says Google 'is and will continue to be' subject to claims involving AI training, where its last annual report said it 'may' be; it names no case and books no reserve.

What each filing says

Microsoft's report filed 28 October lists current copyright claims arising from AI training and says it may have to sign royalty-bearing licenses or pay damages, adding that it has limited visibility into the training data of models it obtains from others. Meta's report filed 22 October says per-work statutory damages could be substantial. Anthropic's IPO filing filed 26 October books the $1.5 billion Bartz settlement, says authors and publishers who opted out of that settlement have sued it individually, and reports demands from more rights holders; it describes its data as public internet data, licensed data, synthetic data and book scanning, and names no pirate library.

What this is not

These are the companies' own risk disclosures to investors, not court findings and not admissions that any particular use was infringing. None of the filings says how any single claim will end.