Character.AI Bankruptcy, Disney’s $20M Bid, Aramco’s Industrial AI Bet, Anthropic’s Cyber Defense, Meta’s Muse Challenge

SPEAKER_1 (0:05)

Welcome to Digimasters Shorts.

SPEAKER_2 (0:07)

We are your hosts, Adam Nagus and Carly Wilson, delivering the latest scoop from the digital realm.

character.ai has filed for Chapter 11 bankruptcy protection in Delaware and plans to sell all or most of its assets through a court-supervised process. The Chatbot platform, founded in 2021, allows users to interact with AI-generated characters. Disney has made a $20 million opening bid for a permanent, non-exclusive license covering some intellectual property and legal claims. Disney will also temporarily provide former Character AI employees to support the platform's safe operation. The offer establishes a minimum price while rival buyers may submit higher bids. character.ai expects the sale to conclude in late November or early December. The company listed estimated assets and liabilities of between $50 million and $100 million.

Its financial difficulties follow lawsuits involving teen mental health crises, suicides and claims brought by families and US states. character.ai says its app and website will continue operating with limited disruption, but the platform could shut down if no buyer supports its continued operation.

SPEAKER_1 (1:13)

Aramco Digital has partnered with startup OrbitronAI to develop and sell artificial intelligence tools for heavy industry. The tools will target oil and gas, power, mining, transport and manufacturing companies. They are designed to reduce losses caused by slow decisions, scattered data and paperwork in major projects and daily operations. The agreement was signed in September during the Leap 2026 technology event in Riyadh, with financial terms undisclosed. OrbitronAI's software uses AI agents to analyse company files, connect with business systems and complete tasks while leaving key approvals and financial decisions to employees. The companies will first develop and test solutions with customers before turning successful projects into products for the wider market. Aramco Digital Chief Executive Ashraf Tahini said the partnership aims to create scalable AI products for Aramco, its affiliates and industrial customers. OrbitronAI Chief Executive Saulius Adamaitis said the companies were building a repeatable model for recovering value across industrial operations. The partnership comes as Aramco increases its AI investments and competes with other startups developing artificial intelligence for supply chains and heavy industry.

Anthropic released OSS Scanner on October 8th, a free AI powered tool designed to find and document vulnerabilities in eligible open source projects. The launch came one week after Google suspended its open source bug bounty submissions, citing an overwhelming number of invalid AI generated reports.

Unlike traditional fuzzing tools, OSS Scanner analyzes code semantics and produces a proof of concept, an explanation of the security impact and a suggested patch when possible. Anthropic says an internal test found 96 real vulnerabilities among 97 reported findings, although independent analysts have not verified the results. The service is available selectively to core maintainers of projects considered critical to infrastructure or user security, including PostgreSQL, OpenSSL, WolfSSL and HotCRP.

Anthropic's disclosure process does not impose a 90-day deadline on validated AI findings, while independently verified vulnerabilities follow the company's standard coordinated disclosure policy. The company also launched its critical infrastructure defense program, pairing Claude models and engineers with security firms, protecting power, water, transportation, and other operational technology networks. Anthropic has not explained how partners will safely test AI-recommended fixes on systems that often cannot be taken offline. Together, the programs represent Anthropic's effort to make artificial intelligence a force for cyber defense as rival companies race to develop similar capabilities.

SPEAKER_2 (3:54)

Meta has introduced MuseWonder and MuseWonder Full, two proprietary AI models designed for complex software engineering, legal work, business automation, and long-running agent tasks. The models come from Meta Superintelligence Labs, led by Alexander Wang, and mark a shift away from the company's earlier focus on downloadable openweight systems.

Independent Evaluations gave Wonderful a score of 56, tying Anthropix Claude Sonnet 5.5 and surpassing OpenAI's GPT-6 Astra, Google's Gemini 4 Argon, and Anthropix Claude Fable 5.1. Meta says the models support text, images, audio and video, offer context windows of about 1 million tokens, and use active memory and active focus to maintain goals and recover from setbacks. The standard Wunder model is available through Meta's API at $4 per million input tokens and $20 per million output tokens, while Wunderful is currently limited to paid MuseCode subscriptions. Neither model can be self-hosted, and Meta has not announced Pay-as-you-go API access for its most capable version.

Enterprise customers must also consider privacy rules, since MuseCode's default contributor mode allows submitted content to help improve future models, although users can disable it. The launch gives businesses another serious frontier AI option, but Meta must still prove that its benchmark results translate into reliable, affordable automation in real-world operations.

SPEAKER_3 (5:16)

Thank you for listening to today's AI and Tech News Podcast summary. Please do leave us a comment, and for additional feedback, please email us at podcast at digimasters.co.uk.