AI News: April 10, 2026

1. Stanford Study Reveals When Multi-Agent AI Systems Are Worth the Compute

Research. A large-scale Stanford study evaluating 180 agent configurations finds that multi-agent AI systems’ apparent advantage largely comes from using more compute, not from the multi-agent architecture itself. However, parallelizable tasks (like financial reasoning) see an 80.9% improvement with centralized coordination, while sequential reasoning tasks suffer when multiple agents are introduced. Source

2. Visa Launches Intelligent Commerce Connect for AI Agent Payments

Payments. Visa debuted Intelligent Commerce Connect (ICC), a platform enabling AI agents to discover merchants, compare prices, and complete purchases on behalf of users using securely accessed card credentials. The platform supports four agentic commerce protocols and works across any card network. Currently in pilot with general availability expected by June 2026. Source

3. Mercor Data Breach Fallout: Lawsuits Mount for $10B AI Startup

Security. AI training-data startup Mercor ($10B valuation) faces escalating consequences from a March supply-chain attack via the LiteLLM open-source library. A hacker group claimed 4TB of stolen data including candidate profiles and API keys. Meta has paused contracts, class-action lawsuits allege 40,000+ people exposed, and the company is losing big-name customers. Source

4. First Conviction Under Take It Down Act Involves AI-Generated Content

Policy. An Ohio man became the first person convicted under the Take It Down Act after pleading guilty to creating AI-generated explicit deepfakes of women and minors using 100+ AI tools. He harassed at least 10 victims and continued generating content even after arrest. The landmark conviction tests the new federal law’s enforcement against AI-enabled non-consensual imagery. Source

5. Sierra Launches Ghostwriter: An Agent for Building Customer Service Agents

Product Launch. Sierra AI’s Ghostwriter lets enterprises create AI customer service agents by uploading SOPs, call transcripts, or whiteboard photos. CEO Bret Taylor argues traditional click-based web applications will be replaced by natural language interfaces powered by autonomous agents. The tool targets enterprise teams who want to deploy agents without engineering resources. Source

6. Factory AI Ships Desktop App for Autonomous Software Development Agents

Developer Tools. Factory released a native desktop app for its “Droid” autonomous AI agents, available on macOS and Windows. Features include persistent machines that retain context between sessions, multi-agent management with simultaneous sessions, direct desktop control (VS Code, browsers, terminals), and local model support via Ollama/vLLM so no data leaves the network. Source

7. Genspark AI Workspace 4.0 Adds Desktop Computer Use and Office Integration

Product Launch. Genspark released AI Workspace 4.0, expanding its “Claw” AI employee to desktop with computer-use capabilities, native Microsoft Office plugins (PowerPoint, Excel, Word), live meeting translation, and a faster execution engine. The company hit $250M ARR in just 12 months, and the release came three weeks after Workspace 3.0. Source

8. HeyGen Launches Avatar V with Identity-Consistent AI Video Generation

Product Launch. HeyGen released Avatar V, which captures a user’s likeness from a 15-second clip and maintains identity consistency across all generated videos regardless of length, angle, outfit, or scene. The system supports 175 languages with automatic lip-sync adaptation, solving the “identity drift” problem at the model level rather than through post-processing. Source

9. Trent AI Emerges from Stealth with $13M to Secure AI Agents

Venture Capital. UK-based Trent AI emerged from stealth with $13M in seed funding led by LocalGlobe and Cambridge Innovation Capital. Founded by former AWS engineers, the startup built a layered security platform to protect AI agents throughout their lifecycle. According to Deloitte’s 2026 report, 75% of companies plan to deploy agentic AI within two years but only 20% have mature governance models. Source

10. Haast Raises $12M Series A for AI-Powered Enterprise Compliance

Venture Capital. New York-based Haast raised $12M in Series A funding led by Peak XV Partners, with participation from DST Global Partners, Airtree, and Aura Ventures. The startup builds compliance infrastructure that uses AI agents to automate regulatory and policy review inside enterprise workflows. Haast reports 4.5x revenue growth in 12 months and zero customer churn among Fortune 500 customers. Source

11. Plus One Robotics Surpasses 2 Billion AI-Powered Warehouse Picks

Robotics. Plus One Robotics announced it surpassed 2 billion successful picks across its global fleet of AI-powered parcel induction and depalletization robots, coinciding with the company’s 10th anniversary. The company doubled its pick count in just two years after taking 8 years to reach the first billion, with operations expanding into Australia and new verticals including wine and food & beverage. Source

12. Jacobi Robotics and ABB Partner on AI-Powered Mixed-Case Palletizing

Robotics. Jacobi Robotics signed a collaboration with ABB Robotics to bring AI-powered mixed-case palletizing to ABB’s integrator network via Jacobi’s OmniPalletizer software. Mixed-case palletizing costs the US industry over $15 billion annually in direct labor alone. The solution will be demonstrated live at MODEX 2026 in Atlanta. Source