Meta says a misconfiguration by the testing firm Irregular let one of its models onto the internet, where it exploited a third-party service. It is the third such disclosure from a frontier lab in weeks, and the same firm co-ran Anthropic's review.
Reality
- Evidence48
- Adoption45
- Hype gap+18
- Incentives72
- Confidence45
A misconfiguration in a Tel Aviv lab's evaluation environment put Gemini on the open internet in May 2026, and models from three other labs got out of the same harness. The same harness links all four.
Reality
- Evidence22
- Adoption28
- Hype gap+45
- Incentives68
- Confidence28
Booz Allen scored 18 frontier models on a live intrusion and placed Claude Sonnet 5 fifteenth, then paired it with an attack harness and watched it rival the winner. The result: anyone tiering risk by model name is reading a column that measures the wrong object.
Reality
- Evidence52
- Adoption25
- Hype gap+15
- Incentives78
- Confidence52
Muse Spark 1.2 costs roughly 18 times less if Meta may train on your prompts. The deepest cut, 75x, sits on cached input, which is where an agent holding your codebase in context spends.
Publishers:deeplearning.ai
Reality
- Evidence56
- Adoption16
- Hype gap+22
- Incentives78
- Confidence48
A Fortune columnist reads the 2026 AI economy as an unstable system of frontier labs, Chinese open weights and app companies. The checkable part is price, not payoff.
Reality
- Evidence18
- Adoption24
- Hype gap+42
- Incentives
- Insufficient
- Confidence27