AI competition risks rise
- Analysts warned Anthropic’s Project Glasswing could create an exclusive consortium for critical‑infrastructure AI work, raising antitrust questions. - The project would centralise vulnerability research and support among a few firms, which critics say could limit competition. - Senator Elizabeth Warren also warned a major AI failure might trigger a financial panic, increasing scrutiny of concentrated AI dependencies ( ).
Warnings about artificial intelligence concentration widened this week as critics said Anthropic’s Project Glasswing could give a small club of companies unusual control over critical-software defense. (promarket.org) Anthropic launched Project Glasswing on April 7 with 12 named partners: Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, Palo Alto Networks, and Anthropic itself. Anthropic said those partners, plus more than 40 additional organizations, would get access to Claude Mythos Preview for defensive cybersecurity work. (anthropic.com) Anthropic said Mythos Preview had already found “thousands” of high-severity vulnerabilities, including flaws in every major operating system and web browser, and said it was committing up to $100 million in usage credits and $4 million in donations to open-source security groups. (anthropic.com) Project Glasswing is built around a simple idea: use a powerful model to scan code for dangerous bugs before attackers do. Anthropic said the target systems include software used in banking, medical records, logistics networks, power grids, and other core services. (anthropic.com) The antitrust warning came on April 22 in ProMarket, where legal scholar Madhavi Singh argued that the arrangement looks like “concerted action” under Section 1 of the Sherman Act. She wrote that sharing sensitive security information inside a closed coalition, while excluding outsiders from the same data and practices, could be treated as an illegal restraint of trade. (promarket.org) Singh’s argument does not say bug-finding partnerships are automatically unlawful. It says the structure matters when direct competitors coordinate, exchange information, and decide who does or does not get access to a tool that could shape the security of widely used infrastructure. (promarket.org) Anthropic has framed the project differently. In its announcement, the company said it would share what it learns so “the whole industry can benefit,” and said governments, software companies, researchers, and open-source maintainers all have roles in the work. (anthropic.com) The concentration debate widened on April 22 when Senator Elizabeth Warren said at a Vanderbilt Policy Accelerator event that a debt-driven artificial intelligence bubble could become a broader financial risk. Warren, the top Democrat on the Senate Banking Committee, said “the parallels to the 2008 financial crisis are striking” and urged “simple structural reforms” before any crash. (publicnow.com) Warren’s warning was about finance, not Glasswing specifically, but it added pressure to questions about what happens when a few firms become central to artificial intelligence tools, infrastructure, and risk management at the same time. (publicnow.com; anthropic.com) For now, Project Glasswing remains both a security program and a test of how far private coordination can go before regulators step in. The next fight is likely to center on whether exclusive access is a temporary safety measure or a durable advantage. (promarket.org; anthropic.com)
