Disney is getting Character.AI’s people and much of its technology, and court filings suggest the legal baggage stays with the bankrupt seller. Whether that structure survives scrutiny from 46 states is the interesting part.
1. Disney gets the team and the tech. Character.AI keeps the lawsuits.
Character.AI filed for Chapter 11 on October 12, and WIRED’s reading of the court filings is that the bankruptcy is the mechanism of the deal, not a sign it failed. Disney has already hired most of the staff, including former CEO Karandeep Anand, who is now its CTO. TechCrunch reports offers to at least 60 people, with an estimated $60 million in annual compensation. A draft court order would say the arrangement is not a merger and that Character.AI stays liable for its obligations. According to a motion cited by WIRED, Disney required the deal to be completed through bankruptcy because of the startup’s litigation and regulatory risks. Disney declined to comment.
What Disney wants is notable: a perpetual, nonexclusive license covering code and tools, including the trust-and-safety moderation stack and age-prediction technology. User data and the website are excluded. Disney paid $5.5 million up front, with a $12.5 million closing payment pending an auction. TechCrunch puts the total at $20 million. It is a stalking-horse bid, so a rival could outbid it.
The catch is that legal shields like this don’t always hold. A judge let a suit against Google proceed on an aiding-and-abetting theory after Google argued it was separate from Character.AI. Forty states have already told the bankruptcy court they reserve the right to challenge any plan. Character.AI disputes their characterization of its platform’s safety.
Ars Technica · TechCrunch · WIRED
2. Exchange admins lose a weekend while a NetScaler flaw demands a fast patch
Two unrelated problems are landing on the same sysadmins. The Register’s headline says one admin spent a whole weekend on his servers after Microsoft’s October updates broke Exchange Server. The supplied excerpt doesn’t include the technical details, so the specifics of the failure remain unverified here.
The other problem is not Microsoft’s. Citrix’s September 27 bulletin reported exploitation of CVE-2026-88771, which allows unauthenticated command execution on NetScaler appliances, including in the default configuration. Further advisories on October 3 and 8 covered SAML issues. Cloudflare, which wrote the explainer we have, shipped an emergency WAF rule on October 1.
The shared lesson is the old dilemma: patch fast and risk an outage, or wait and risk compromise. Cloudflare’s answer is to put access controls in front of login pages, which is also what Cloudflare sells. That doesn’t make the advice wrong. A gateway that must be reachable by everyone is a bigger target than one that checks identity first. But it doesn’t help the admin whose Exchange server just broke.
The Register · Cloudflare Blog
3. Microsoft reportedly drops a UK AI supercomputer as the grid falls short
The Register’s headline says Microsoft has abandoned a UK AI supercomputer as the grid struggles to keep up. The text we have is just that headline, so details like the site, the connection timeline, and Microsoft’s own explanation are not confirmed here.
Still, the constraint it names is a physical one that no model improvement fixes. A nearby Register item says TSMC is sending scouts to check whether American datacenters can deliver, with up to $64 billion budgeted for expansion while it watches construction progress and power supplies. If chipmakers are auditing the electricity as well as the buildings, power is part of the investment case, not a footnote.
4. A Broadcom Wi-Fi bug turns 16 and still reportedly hijacks routers
The Register’s headline says a 16-year-old Broadcom Wi-Fi bug still lets strangers hijack your router. That is all the supplied text tells us. We don’t have the technical details, the affected products, or Broadcom’s response, so claims about how many devices are exposed or whether they can be patched would be guesswork.
The age alone is the uncomfortable part. A flaw old enough to drive predates much of today’s security tooling, and consumer networking gear is rarely updated once it ships. If the report holds up, the question for anyone with a router is whether the vendor ever shipped, or can still ship, a fix for it.
5. ConnectWise switches off cloud services while it investigates an incident
ConnectWise has turned off its SaaS backup, security, and management services while it probes what The Register calls a security SNAFU. The supplied text is only that headline, so the cause, the number of affected customers, and any timeline for restoration are unknown.
The decision itself says something. Shutting down services that customers depend on for backup and security is expensive and disruptive, and companies rarely do it unless they want to contain something. For the managed-service providers who run client environments through these tools, the outage is now their problem too. The open question is how much ConnectWise will disclose before bringing the services back.
