AI News from August 19, 2026

Cognition CEO Scott Wu says SpaceX never pursued a deal, challenging a Bloomberg report as SpaceX expands its AI coding and enterprise strategy.

OpenAI is previewing Private Safety Processing for Zero Data Retention customers, aiming to detect multi-session misuse without exposing prompts to staff.

NVIDIA released SkillEvaluator, an open-source framework that tests whether agent skills improve task results, safety, and efficiency across real runs.

Coinbase’s Base is backing AI-agent startups with a $100,000 accelerator, signaling a push to bring autonomous software into crypto applications.

Netwrix is extending identity security monitoring to AI agents in Microsoft Entra ID, signaling a shift toward governing nonhuman access.

TrueFoundry has launched TrueForge, an open-source, vendor-neutral alternative to Claude Managed Agents that it says can cut costs by 50%.

Bridgewater investors are backing an AI startup aimed at giving smaller hedge funds stronger research tools, signaling broader access to AI.

Baidu shares fell after earnings despite reported AI revenue growth and cloud expansion, highlighting investor concerns over margins, guidance, and monetization.

Alibaba has released a laptop-focused AI model ahead of Meta’s expected move, sharpening competition over capable, locally run AI tools for builders.

Block’s free Goose agent offers local coding automation without subscriptions, giving developers a privacy-focused alternative to Claude Code’s paid limits.

Baidu reports a fifth consecutive quarterly revenue decline, sharpening questions about whether its AI strategy can reverse pressure from rivals in China.

Baidu CEO Robin Li says Ernie will return to the AI frontier after a revenue miss and 68% profit drop sent shares lower, raising execution questions.