• Early April: Mythos is unveiled but held back
• Project Glasswing: From discovery to global defense
• Government and religious responses: Policy “adolescence”
• Intelligence use and legal friction

Anthropic’s ‘Mythos’ AI Model Triggers Government and Religious Responses Anthropic’s decision to build—but not release—its powerful “Mythos” AI model has jolted governments, security agencies, and ethicists into a new phase of AI anxiety, where defensive safeguards and offensive capabilities are advancing at the same time.

§Early April: Mythos is unveiled but held back

In early April, Anthropic announced Mythos, a general-purpose AI system “particularly capable at computer security tasks,” yet chose to keep it unreleased to the public. During internal testing, the company said Mythos Preview could “identify and exploit vulnerabilities in every major operating system and every major web browser,” even uncovering a 27-year‑old bug in an OS “known primarily for its security.”

§Project Glasswing: From discovery to global defense

Following the announcement, Anthropic launched Project Glasswing with 50 initial partners—companies, NGOs, labs, and the U.S. government—to harden digital infrastructure. Using Mythos, participants reportedly discovered over 10,000 high‑ or critical‑severity flaws, many in infrastructure whose compromise “could be affected, with potential ramifications for global and national security.”

Five days before the newsletter’s publication, Anthropic expanded Glasswing to more than 150 organizations across 15 countries, prioritizing operators of vital services. The company argues Mythos has already “transformed the current cybersecurity paradigm,” and predicts that “other AI companies will have Mythos‑class models within 6 to 12 months,” potentially without comparable safeguards.

§Government and religious responses: Policy “adolescence”

These rapid advances, and the specter of AI‑driven cyberattacks on critical infrastructure, have contributed to what one observer calls the “adolescence of AI policy,” marked by a U.S. executive order pushing voluntary security frameworks and a papal encyclical urging tighter regulation and multilateral governance.

§Intelligence use and legal friction

Even as policymakers warn of misuse, intelligence services are reportedly embracing Mythos. The U.S. National Security Agency is alleged to be “using Anthropic’s Mythos for cyber attacks,” an arrangement emerging while the lab is “locked in [a] legal battle with [the] Pentagon over [its] Claude model.” This juxtaposition—defensive partnerships, offensive use by a spy agency, and courtroom fights over AI deployment—illustrates the increasingly tangled, global contest over how far such systems should go.

Continue reading https://foxvector.com/stories/019ea64c-0c1d-16e8-716e-062916aa1a1c