This story needed a sharper line between what is confirmed and what is rumor
The original version mixed one very real announcement with one rumor-heavy narrative. Anthropic has now officially published extensive material on Claude Mythos Preview and Project Glasswing. OpenAI, meanwhile, has publicly announced GPT-5.4 and continued Sora-related product activity. What OpenAI has not publicly confirmed is a product called GPT-5.5 "Spud".
That distinction matters. If this site is going to be useful, it needs to separate frontier-model gossip from what companies have actually put on the record.
What Anthropic has officially confirmed
Anthropic's Project Glasswing announcement is not subtle. The company says Mythos Preview is an unreleased frontier model that can outperform all but the most skilled humans at finding and exploiting software vulnerabilities. Anthropic also says Mythos Preview has already found thousands of high-severity vulnerabilities, including issues across every major operating system and major web browser.
Rather than releasing that model broadly, Anthropic says it is giving controlled access to partners through Project Glasswing so those capabilities can be used defensively. The public announcement names major launch partners including AWS, Apple, Cisco, CrowdStrike, Google, JPMorganChase, Microsoft, NVIDIA, and Palo Alto Networks, and says Anthropic is committing up to $100 million in usage credits.
That is already a major story on its own. It signals that frontier labs now believe the cyber capability jump is large enough to justify limited release and coordinated defensive deployment.
What OpenAI has officially confirmed
OpenAI's latest clearly documented frontier release is GPT-5.4, announced on March 5, 2026 for ChatGPT, the API, and Codex. OpenAI describes it as its most capable and efficient frontier model for professional work.
OpenAI has also continued to discuss Sora in official channels. In late 2025 it announced a major Disney licensing agreement tied to Sora, and OpenAI's own support materials still documented Sora changes and discontinuation details in 2026. That does not line up with the earlier version of this article, which presented Sora as something OpenAI had effectively sacrificed in order to build an unconfirmed model.
What is not confirmed
There is no official OpenAI announcement, product page, or documentation page that publicly confirms a released or imminent model called GPT-5.5 "Spud". That does not prove internal codenames do not exist, but it does mean the previous version of this article leaned too heavily on rumor and secondary reporting.
For now, the careful way to say this is: Anthropic's Mythos story is official; the public Spud story is not.
Why this still matters
The bigger pattern remains real even after the correction. Frontier AI companies are moving in two visible directions at once:
- More capable agentic systems for coding, reasoning, and professional work
- More caution around capabilities that create obvious misuse risk, especially in cybersecurity
Anthropic's stance is the clearest example of the second trend right now. OpenAI's confirmed releases point more toward broad deployment of increasingly capable general-purpose tools.
The corrected takeaway
If you want the verified story in April 2026, it is this: Anthropic has made a consequential public move with Mythos Preview and Project Glasswing. OpenAI's current public record points to GPT-5.4, Codex, and ongoing Sora-related activity. Anything more specific about a rumored GPT-5.5 "Spud" should be treated as unconfirmed unless and until OpenAI publishes it.
That is a less dramatic headline than the original article, but it is a more trustworthy one.
Cover image attribution: official screenshot captured from Anthropic's Project Glasswing page on 2026-04-08: Anthropic: Project Glasswing.
Sources used for this correction:
