The AI Dispatch

Frontier Models

GPT-5.5 “Spud” Nears Release — Polymarket Puts 78% Odds on April Launch

OpenAI’s next model completed pretraining on March 24 and prediction markets are betting heavily that it ships before the month is out. The question is no longer whether, but what name it wears when it arrives.

Sources: Panstag

OpenAI’s next frontier model, internally codenamed “Spud,” completed its pretraining run on March 24 and has entered the safety evaluation and red-teaming phase that typically precedes a public release. A Polymarket contract tracking whether the model ships before April 30 currently trades at seventy-eight cents — reflecting strong consensus among prediction-market participants that a launch is imminent. Sam Altman described the model as “very strong” and suggested it could “really accelerate the economy,” while co-founder Greg Brockman characterized it as the culmination of “two years of research.”

The naming remains an open question. Whether it ships as GPT-5.5 or leapfrogs to GPT-6 will depend on how its benchmark performance compares to the existing GPT-5 series. Industry observers note that the distinction matters less for capability than for marketing: the version number signals to enterprise buyers and developers how much of a generational jump to expect, and OpenAI has historically been deliberate about when it increments the major version. Regardless of the label, the model is expected to represent a significant step forward in reasoning, multilingual fluency, and agentic task completion.

The timing is notable. With Anthropic’s Mythos revelations dominating security headlines this week and DeepSeek V4 still rolling out, a Spud release would ensure that no single lab dominates the narrative for long. The frontier-model release cadence has compressed from roughly annual to quarterly, and the market’s confidence in an April launch suggests that cadence is still accelerating.

Copyright & Litigation

Anthropic Copyright Settlement Update — Six Authors Opt Out, File Individual Suits

The largest AI copyright settlement in history hits a procedural snag as six authors reject the class deal and pursue individual claims against every major lab, seeking $150,000 per title per defendant.

Sources: Writer Beware

The Authors Guild published its April update on the $1.5 billion Bartz v. Anthropic class-action settlement — the largest AI copyright settlement to date — and the news is mixed. While the vast majority of class members have not opted out, six authors have formally rejected the settlement and filed individual lawsuits against Anthropic, OpenAI, Google, Meta, xAI, and Perplexity AI. Each plaintiff is seeking $150,000 per copyrighted title per defendant, a damages framework that could result in payouts far exceeding what the class settlement offers on a per-author basis.

The opt-outs are strategically significant even if they represent a small fraction of the class. Individual suits allow plaintiffs to pursue discovery against each defendant separately, potentially unearthing details about training data practices that a class settlement would have kept sealed. The final approval hearing for the Bartz settlement is scheduled for May 14, and the presiding judge will weigh whether the opt-out rate is low enough to proceed. Legal analysts expect approval, but the individual suits will continue regardless, creating a parallel litigation track that could produce binding precedent on whether training on copyrighted text constitutes fair use.