AI Dev News Digest: March 6th, 2026 (www.everydev.ai)

Anthropic recently declined a military contract with the Pentagon, causing significant ripples across the AI landscape. This decision correlates to Dario Amodei's leaked memo, which criticized OpenAI's military dealings as merely "safety theater." Amidst this controversy, Anthropic's user engagement surged, with Claude now attracting over a million daily signups. Their new feature, Import Memory, facilitates users' transition from ChatGPT to Claude by allowing easy memory transfers. These developments highlight a critical ecosystem shift as developers rally around companies that prioritize ethical AI use. In response to competitive pressures, OpenAI accelerated its own advancements, launching GPT-5.3 Instant and GPT-5.4, with improvements like a 1M token context and enhanced desktop processing capabilities. OpenAI also introduced Codex Security, designed to better identify and manage project-specific vulnerabilities, which flagged thousands of security issues across major codebases. The AI security arms race intensified, as both organizations discovered and reported vulnerabilities: Anthropic's Claude identified 22 zero-days in Firefox, while OpenAI scanned 1.2 million commits for bugs, underscoring the urgent need for robust security in AI implementations. This week's news reflects an escalating rivalry focused not only on technological advancements but also on principles of ethical engagement and security in AI development.