Senior AI Journalist
Thinking Machines Lab Seeks $1 Billion Funding at $40 Billion Valuation, Signaling Continued AI Investment Boom
In a move that underscores the persistent investor confidence in artificial intelligence, Thinking Machines, the AI research lab co-founded by former OpenAI CTO Mira Murati, is reportedly in discussions to secure a substantial $1 billion funding round. This new investment would catapult the company’s valuation to an impressive $40 billion. The Information reported on these talks, indicating that existing backer Accel is a likely candidate to lead this significant capital injection. This development comes despite the company reportedly seeking a higher $50 billion valuation late last year, suggesting a slight recalibration but still reflecting a robust market appetite for groundbreaking AI ventures.
Thinking Machines, which launched in early 2025, has rapidly ascended in the competitive AI landscape. Their success is partly attributed to the strong pedigree of its founders, drawing talent from top-tier organizations like OpenAI. The proposed $40 billion valuation, against an annual revenue run rate exceeding $100 million, highlights an exceptionally high revenue multiple, a common characteristic in the high-growth AI sector where potential often outweighs immediate profitability. This valuation speaks volumes about the perceived long-term impact and market dominance investors anticipate from Thinking Machines, particularly following their introduction of Inkling, an open-weight model that leverages usage-based compute fees on its Tinker platform.
The company’s previous funding round was a massive $2 billion seed financing, one of the largest in history, led by Andreessen Horowitz with participation from industry giants like Nvidia, GV, Lightspeed, and Conviction Partners. Such substantial backing from inception indicates a deep belief in the team’s vision and technological prowess. However, the journey hasn’t been without its challenges, with some high-profile co-founders, including Lilian Weng and Luke Metz, reportedly returning to OpenAI. This dynamic flow of talent between leading AI labs underscores the intense competition and the constant pursuit of innovation at the very apex of artificial intelligence research and development. The new funding, if secured, will further empower Thinking Machines to push the boundaries of AI research, attract top talent, and expand its technological offerings, potentially reshaping various industries with its advanced AI solutions.
Source: TechCrunch
Abliteration.ai Commercializes AI Guardrail Removal, Raising Ethical Concerns and Cybersecurity Debates
A new startup, Abliteration.ai, is carving out a controversial niche in the artificial intelligence industry by offering a service that removes the built-in guardrails and refusal mechanisms from open-weight AI models. This practice, known as “abliteration,” allows users to access highly capable AI models, such as Z.ai’s GLM-5.3, without the ethical restrictions designed to prevent harmful outputs. The company, officially incorporated in March 2026, aims to democratize access to these uncensored models, facilitating applications in offensive cyber operations, red-teaming, and agent testing that other models might refuse to perform.
While the concept of removing guardrails from open-source models is not new—researchers and developers have long engaged in this practice—Abliteration.ai’s innovation lies in commercializing and simplifying access to these models. By providing a web browser interface and API access, Abliteration.ai eliminates the significant friction involved in downloading, configuring, and computing these models locally. TechCrunch demonstrated the ease of use, successfully prompting an abliterated GLM-5.3 to generate Python code for stealing Chrome passwords and a protocol for culturing dangerous human pathogens, tasks typically refused by standard AI models.
The company’s founders argue that enabling access to uncensored frontier models is crucial for cybersecurity defense. Their rationale is that defenders need to understand and replicate the behaviors of malicious actors to effectively protect against them. Devon, a co-founder whose full name is withheld due to his employment at another firm, states that abliterated models help defenders “model bad actors” and “accelerate cybersecurity.” Abliteration.ai reportedly serves several early-stage red-teaming startups in the UK and Europe, which work with critical infrastructure providers like banks and airlines to enhance their cybersecurity postures. These customers, according to Devon, would be unable to use standard, guarded AI models for their specific red-teaming needs.
However, the emergence of Abliteration.ai has ignited serious ethical concerns among AI safety experts. Andrew Yoon, head of research at CivAI, expressed alarm to TechCrunch, describing the process as “modify[ing] the model so that it becomes a sociopath.” Critics fear that making such powerful, unrestricted models widely accessible could lead to significant harm, predicting an increase in their misuse for malicious purposes in the near future. The debate extends to the efficacy of abliterated models, with some cybersecurity experts like Ahmed Aly of Fabraix suggesting that the abliteration process might degrade the model’s overall knowledge and capabilities, potentially making them less effective for sophisticated malicious acts compared to fine-tuned, less restrictive open models. Nevertheless, the platform is challenging the industry to confront fundamental questions about responsibility and the boundaries of AI accessibility.
Source: TechCrunch
US Government Backs OpenAI in Copyright Dispute, Prioritizing AI Innovation Over Creator Protections
In a significant legal development, the U.S. government has filed an amicus brief in support of OpenAI, the maker of ChatGPT, in a lawsuit brought by The New York Times. The brief defends OpenAI’s practice of training its large language models (LLMs) on copyrighted material without explicit permission from creators. This intervention by the Trump administration underscores a clear prioritization of fostering a robust and competitive artificial intelligence industry within the United States, positioning it as critical for global leadership in AI technology.
The core of the legal battle revolves around the concept of “fair use,” a doctrine within copyright law that allows for the unlicensed use of copyrighted works under specific circumstances. The government’s brief argues that constraining LLM development due to a “misunderstanding of fair use doctrine” would impede creative and scientific progress, ultimately hindering American prosperity and economic mobility. This perspective suggests that the transformative nature of AI, which utilizes vast datasets to generate new content, should be considered distinct from traditional copyright infringement, akin to how a human learns from existing works before creating their own.
The lawsuit by The New York Times is one of many challenging AI companies for their use of copyrighted material. Publishers argue that AI models are trained on their content, effectively devaluing their intellectual property without fair compensation. However, past legal precedents in AI training and copyright have often leaned in favor of AI developers, with the notable exception of cases involving the use of illegally obtained material. For instance, a $1.5 billion settlement against Anthropic was for using “shadow libraries” to pirate books, not for the act of training AI on copyrighted content itself.
While the amicus brief from the Trump administration is not a binding ruling, it carries substantial weight in the ongoing legal discourse. It signals the government’s stance that the benefits of rapid AI development, including potential advancements in various sectors, outweigh the immediate concerns of copyright holders regarding data ingestion for training. This strategic alignment aims to accelerate AI innovation, but it also deepens the divide between AI developers and content creators, raising fundamental questions about the future of intellectual property in the age of artificial intelligence and how best to balance innovation with fair compensation for original works. The ultimate resolution of these cases will likely set critical precedents for the future landscape of AI development and content creation.
Source: TechCrunch
This article was produced with the assistance of AI tools and reviewed by the AIStackDigest editorial team.