AI News Roundup — April 11, 2026
Anthropic launches a major cybersecurity initiative with a $100 million credit commitment, Google releases the open-source Gemma 4 model family, and OpenAI surpasses $25 billion in annualized revenue as it eyes a public listing. Meanwhile OpenAI, Anthropic, and Google form a rare joint coalition to combat unauthorized model distillation.
Anthropic Launches Project Glasswing Cybersecurity Initiative
Anthropic unveiled Project Glasswing, granting CrowdStrike, Microsoft, Apple, and Google access to a Claude Mythos2 preview model for defensive security work. The model already uncovered a 27-year-old bug in OpenBSD. Anthropic also committed $100 million in credits to fix critical vulnerabilities before bad actors can exploit them.
Google Releases Gemma 4 Open Model Family
Google launched Gemma 4 under an Apache 2.0 license — its most capable open model family to date. Available in four sizes with native vision, audio, function calling, a 256K context window, and support for 140+ languages, the release significantly raises the bar for open-source AI accessibility.
OpenAI Surpasses $25 Billion ARR, Eyes IPO
OpenAI closed its $122 billion funding round at an $852 billion valuation and has surpassed $25 billion in annualized revenue. The company is reportedly taking early steps toward a public listing, potentially as soon as late 2026. Rival Anthropic is closing in at $19 billion in annualized revenue.
Meta Confirms Muse Spark as Proprietary, Plans $115–135 Billion AI Capex
Meta confirmed its new Muse Spark model will remain proprietary, marking a strategic shift away from its open-source roots. The company also disclosed its 2026 AI capital expenditure will total $115–135 billion — nearly double last year — representing one of the largest single-year AI infrastructure bets in corporate history.
OpenAI Proposes Robot Taxes, Public Wealth Fund, and Four-Day Workweek
OpenAI published a 13-page policy paper outlining its vision for managing AI-driven economic disruption. Proposals include taxes on robots, a government-managed public wealth fund, and a four-day workweek — framing AI companies as active stakeholders in labor policy outcomes rather than passive bystanders.
OpenAI, Anthropic, and Google Form Joint Anti-Distillation Coalition
The three AI giants announced a joint initiative to detect and counter unauthorized model distillation — the practice of training competing models by feeding them outputs from proprietary systems. The coalition represents a rare moment of cooperation among fierce rivals on a shared intellectual property protection concern.
Anthropic Launches Advisor Strategy for Messages API
Anthropic introduced a new Advisor Strategy that pairs Claude Opus 4.6 as a high-reasoning advisor alongside faster Sonnet or Haiku models in the Messages API. The pattern lets developers get best-in-class reasoning on complex subtasks without paying Opus-level compute costs for every inference call.