Anthropic faces antitrust questions

Anthropic’s new cyber-defense project is drawing antitrust scrutiny because it gives a handpicked group early access to Claude Mythos Preview, its unreleased security model. (anthropic.com) (promarket.org) Anthropic announced Project Glasswing on April 7, 2026, with 12 named launch partners: Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, Palo Alto Networks, and Anthropic itself. The company said it has also extended access to more than 40 additional organizations that build or maintain critical software infrastructure. (anthropic.com) The company said Claude Mythos Preview has found thousands of high-severity vulnerabilities, including flaws in every major operating system and web browser, and that Anthropic is committing up to $100 million in usage credits plus $4 million in donations to open-source security groups. (anthropic.com) (red.anthropic.com) A software vulnerability is a bug that can let an attacker break in, steal data, or shut systems down. Anthropic said Mythos Preview can identify and exploit zero-day flaws, meaning previously unknown bugs, and can also turn known-but-unpatched weaknesses into working attacks. (red.anthropic.com) (anthropic.com) Anthropic said it kept Mythos Preview out of general release because the model’s cyber capabilities could be misused, and framed Glasswing as a defensive response to that risk. The company also said it plans to share what it learns so “the whole industry can benefit.” (anthropic.com 1) (anthropic.com 2) The antitrust question starts with who gets in. In a commentary published April 22, Madhavi Singh wrote in ProMarket that an exclusive consortium of major companies sharing sensitive security information and excluding outsiders could raise issues under Section 1 of the Sherman Act, the U.S. law that bars certain restraints of trade. (promarket.org) That argument lands at a moment when U.S. enforcers are already reexamining collaborations among competitors. The Federal Trade Commission and Justice Department withdrew their 2000 collaboration guidelines in December 2024, and in February 2026 opened a joint public inquiry on new guidance for competitor collaborations, including alliances involving new technologies and data sharing. (ftc.gov) (justice.gov) Federal regulators have also been looking closely at artificial-intelligence tie-ups involving Anthropic. On January 17, 2025, the Federal Trade Commission issued a staff report on partnerships and investments linking Alphabet, Amazon, and Microsoft with Anthropic and OpenAI, saying those deals could affect access, influence, and competition in generative AI markets. (ftc.gov) Glasswing’s backers say the point is speed, not market power. Cisco’s Anthony Grieco said the work is “too important and too urgent to do alone,” and Senator Mark Warner of Virginia said the partner companies were “proactively sharing information, capabilities, and computing capacity” to protect critical infrastructure. (anthropic.com) (warner.senate.gov) The pressure on Anthropic increased this week after reports that unauthorized users gained access to Mythos through a third-party vendor environment on the day the model was revealed. Anthropic said it is investigating that possible breach, a reminder that the company is trying to control both who can use the model and how securely that control can be maintained. (msn.com) (thenextweb.com)