Project Glasswing's allowlist is Amazon, Apple, Microsoft — zero audit firms, zero protocol foundations. Mythos surfaced 271 Firefox vulns in supervised conditions on one curated codebase; the downstream effect is whatever happens when this model class hits public API tiers and gets pointed at 5K lines of Solidity. Immunefi has cleared $100M+ in lifetime payouts, and that's the entire whitehat incentive pool against attackers who no longer need PhDs in vuln research.

Top comment by @Benthic

More on Cybersecurity

Amid rising AI-driven cyber threats, the Trump administration’s staff and budget cuts have weakened CISA and sidelined it from the White House’s central AI cybersecurity planning role.

Amid rising AI-driven cyber threats, the Trump administration’s staff and budget cuts have weakened CISA and sidelined it from the White House’s central AI cybersecurity planning role.

Axios ·

Yield Basis selected Firepan to perform an AI-powered security review of its live mainnet FeeDistributor contract, identifying 18 findings across 22 attack surfaces, including a previously undocumented MEV vector.

Yield Basis selected Firepan to perform an AI-powered security review of its live mainnet FeeDistributor contract, identifying 18 findings across 22 attack surfaces, including a previously undocumented MEV vector.

Firepan ·

OpenAI expands cybersecurity push with GPT-5.4-Cyber access for US and UK agencies, partnering with banks and tech giants to strengthen global AI defense systems

OpenAI expands cybersecurity push with GPT-5.4-Cyber access for US and UK agencies, partnering with banks and tech giants to strengthen global AI defense systems

𝕏/@OpenAINewsroom ·

NVIDIA CEO Jensen Huang says China already has the compute capacity to train Claude Mythos-level AI, raising cybersecurity concerns

NVIDIA CEO Jensen Huang says China already has the compute capacity to train Claude Mythos-level AI, raising cybersecurity concerns

CoinTelegraph ·

Anthropic Mythos leaked via unsecured cache before announcement — company says it poses unprecedented cybersecurity risks

Anthropic Mythos leaked via unsecured cache before announcement — company says it poses unprecedented cybersecurity risks

Fortune ·

Leak exposes Anthropic’s Claude Mythos model as company confirms testing of its most advanced AI, warning of serious cybersecurity implications

Leak exposes Anthropic’s Claude Mythos model as company confirms testing of its most advanced AI, warning of serious cybersecurity implications

𝕏/@Cointelegraph ·

Explore the full feed

Comments

Up nextZcash developers propose Ironwood, a network upgrade designed to restore trustless verification of ZEC supply after the Orchard vulnerability exposed risks around undetectable counterfeiting