Research
Aug 18, 2026
AI Context Layers Expose Agent Failures, Not Prevent Them
A VB Pulse July 2026 survey of 101 enterprises finds that those running or building a governed AI context layer report recurring agent failures at 50% — more than twice the rate of those without one. The paradox is deliberate: a governed layer makes failures visible rather than causing them. The data reveals a growing gap between where enterprises are spending and where the problem is actually being solved.
Enterprise AI
AI Agents
Data Governance
Aug 6, 2026
Jeff Dean, Top Google AI Researchers Launch New Startup
Jeff Dean, one of the most storied figures in AI research, is leaving Google alongside several senior colleagues to found a new startup focused on accelerating scientific discovery through AI. The departure marks one of the most significant talent exits from Google's AI division in recent memory.
AI
Startups
Research
Jul 31, 2026
Why VC-Backed Startups Commit More Fraud, Study Finds
A joint study from Imperial College London and Emlyon Business School digs into the mechanics of startup fraud, finding that venture backing itself may create conditions that make misconduct more likely. Researchers mapped how founder incentives, investor pressure, and governance gaps combine to elevate fraud risk. Here's what the findings mean for the startup ecosystem.
Venture Capital
Startups
Fraud
Jul 30, 2026
Anthropic's Mythos AI Finds Long-Hidden Crypto Flaws
Anthropic's Mythos security model has identified incremental weaknesses in the mathematical foundations of two cryptographic algorithms. The findings don't break any real-world systems today, but they signal a potentially important shift in how AI could accelerate cryptanalysis research.
AI
Cybersecurity
Cryptography
Jul 18, 2026
Defenders Weaponize Prompt Injection to Stop AI Hackers
Security researchers at Tracebit have discovered that the same prompt injection technique attackers use to hijack AI systems can be flipped into a powerful defense. By planting malicious-looking prompts alongside cloud secrets, defenders can trigger an AI agent's own safety guardrails — shutting it down cold before it causes serious damage.
AI
Cybersecurity
LLMs