Anthropic antitrust warning
What happened
- ProMarket warned that Anthropic's Project Glasswing could pose antitrust risks if access to Claude Mythos is concentrated. - Project Glasswing is described as an exclusive consortium using Claude Mythos to identify and fix vulnerabilities in critical digital infrastructure. - The analysis argues exclusive access to general-purpose AI capabilities could trigger regulatory scrutiny over market power (promarket.org).
Why it matters
A competition warning is now hanging over Anthropic’s Project Glasswing, after ProMarket argued on April 22 that the closed security consortium could draw antitrust scrutiny. (promarket.org) Anthropic launched Project Glasswing on April 7 and said it would give early access to Claude Mythos Preview to a set of launch partners including Amazon Web Services, Apple, Cisco, Google, JPMorganChase, Microsoft, NVIDIA, and Palo Alto Networks. Anthropic said it has also extended access to more than 40 additional organizations and committed up to $100 million in usage credits plus $4 million in donations to open-source security groups. (anthropic.com) Anthropic said the program is aimed at “the world’s most critical software” and that partners are using Mythos Preview for defensive security work on code and infrastructure used by billions of people. The company said it plans to share what it learns with the broader industry. (anthropic.com) The antitrust issue starts with how a closed group works. ProMarket’s Madhavi Singh wrote that Glasswing is a “concerted action” and argued that information-sharing inside the group, or excluding outsiders from security data and best practices, could be treated as a restraint of trade under Section 1 of the Sherman Act. (promarket.org) That argument lands in a market where access to top-tier general-purpose AI models is already concentrated among a few labs and cloud companies. Singh wrote that if one company can decide which firms get a powerful model for cybersecurity and which firms do not, regulators could ask whether control over that capability is becoming market power. (promarket.org) Anthropic’s case for restricting access is that Mythos is not a normal product launch. In its system card dated April 7, the company said Mythos Preview can identify and exploit zero-day vulnerabilities across every major operating system and every major web browser, and said it does not plan to make the model generally available. (anthropic.com) Anthropic also said those cyber capabilities were not narrowly trained for one task but emerged from broader gains in coding, reasoning, and autonomy. That detail matters for competition questions because ProMarket’s critique is about access to a general-purpose model, not just a specialized security tool. (anthropic.com; promarket.org) Supporters of the project are framing it as a defensive response to a new cyber risk. Senator Mark Warner said on April 7 that AI is already helping threat actors improve their capabilities and said industry should “accelerate and reprioritize patching” as AI speeds up vulnerability discovery. (warner.senate.gov) Legal analysts outside Anthropic are also reading the move as a safety-driven restriction. A National Law Review article published April 22 said Anthropic chose Glasswing instead of a broad release because of the cybersecurity risks of wider deployment and noted that federal financial regulators are already engaging bank chief executives on frontier-model capabilities. (natlawreview.com) The next question is whether Anthropic can keep Glasswing both closed and broadly beneficial. Its own announcement promises industry-wide spillovers, while the antitrust warning says regulators may test whether those benefits are reaching firms outside the club. (anthropic.com; promarket.org)
Key numbers
- A competition warning is now hanging over Anthropic’s Project Glasswing, after ProMarket argued on April 22 that the closed security consortium could draw antitrust scrutiny.
- Anthropic said it has also extended access to more than 40 additional organizations and committed up to $100 million in usage credits plus $4 million in donations to open-source security groups.
- In its system card dated April 7, the company said Mythos Preview can identify and exploit zero-day vulnerabilities across every major operating system and every major web browser, and said it does not plan to make the model generally available.
- Senator Mark Warner said on April 7 that AI is already helping threat actors improve their capabilities and said industry should “accelerate and reprioritize patching” as AI speeds up vulnerability discovery.
What happens next
- A competition warning is now hanging over Anthropic’s Project Glasswing, after ProMarket argued on April 22 that the closed security consortium could draw antitrust scrutiny.
- The company said it plans to share what it learns with the broader industry.
- Singh wrote that if one company can decide which firms get a powerful model for cybersecurity and which firms do not, regulators could ask whether control over that capability is becoming market power.
Quick answers
What happened in Anthropic antitrust warning?
ProMarket warned that Anthropic's Project Glasswing could pose antitrust risks if access to Claude Mythos is concentrated. Project Glasswing is described as an exclusive consortium using Claude Mythos to identify and fix vulnerabilities in critical digital infrastructure. The analysis argues exclusive access to general-purpose AI capabilities could trigger regulatory scrutiny over market power (promarket.org).
Why does Anthropic antitrust warning matter?
A competition warning is now hanging over Anthropic’s Project Glasswing, after ProMarket argued on April 22 that the closed security consortium could draw antitrust scrutiny. (promarket.org) Anthropic launched Project Glasswing on April 7 and said it would give early access to Claude Mythos Preview to a set of launch partners including Amazon Web Services, Apple, Cisco, Google, JPMorganChase, Microsoft, NVIDIA, and Palo Alto Networks. Anthropic said it has also extended access to more than 40 additional organizations and committed up to $100 million in usage credits plus $4 million in donations to open-source security groups. (anthropic.com) Anthropic said the program is aimed at “the world’s most critical software” and that partners are using Mythos Preview for defensive security work on code and infrastructure used by billions of people. The company said it plans to share what it learns with the broader industry. (anthropic.com) The antitrust issue starts with how a closed group works. ProMarket’s Madhavi Singh wrote that Glasswing is a “concerted action” and argued that information-sharing inside the group, or excluding outsiders from security data and best practices, could be treated as a restraint of trade under Section 1 of the Sherman Act. (promarket.org) That argument lands in a market where access to top-tier general-purpose AI models is already concentrated among a few labs and cloud companies. Singh wrote that if one company can decide which firms get a powerful model for cybersecurity and which firms do not, regulators could ask whether control over that capability is becoming market power. (promarket.org) Anthropic’s case for restricting access is that Mythos is not a normal product launch. In its system card dated April 7, the company said Mythos Preview can identify and exploit zero-day vulnerabilities across every major operating system and every major web browser, and said it does not plan to make the model generally available. (anthropic.com) Anthropic also said those cyber capabilities were not narrowly trained for one task but emerged from broader gains in coding, reasoning, and autonomy. That detail matters for competition questions because ProMarket’s critique is about access to a general-purpose model, not just a specialized security tool. (anthropic.com; promarket.org) Supporters of the project are framing it as a defensive response to a new cyber risk. Senator Mark Warner said on April 7 that AI is already helping threat actors improve their capabilities and said industry should “accelerate and reprioritize patching” as AI speeds up vulnerability discovery. (warner.senate.gov) Legal analysts outside Anthropic are also reading the move as a safety-driven restriction. A National Law Review article published April 22 said Anthropic chose Glasswing instead of a broad release because of the cybersecurity risks of wider deployment and noted that federal financial regulators are already engaging bank chief executives on frontier-model capabilities. (natlawreview.com) The next question is whether Anthropic can keep Glasswing both closed and broadly beneficial. Its own announcement promises industry-wide spillovers, while the antitrust warning says regulators may test whether those benefits are reaching firms outside the club. (anthropic.com; promarket.org)
