Air Security's Plugin4Shell lets a plugin be swapped during a background refresh while the agent reports the audited commit. Anthropic and OpenAI have patched, Copilot has no fix, and Google is retiring Gemini CLI.
Reality
- Evidence60
- Adoption65
- Hype gap+20
- Incentives70
- Confidence58
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
Johann Rehberger got code execution in up to 80% of attempts by making one tool fail and letting the agent improvise. Anthropic says Auto Mode is working as designed, which leaves the sandbox decision on your desk.
Reality
- Evidence63
- Adoption70
- Hype gap−8
- Incentives58
- Confidence66
Wiz withdrew the load-bearing part of its Snowflake finding after GitHub's review. The defect underneath is ordinary workflow syntax: untrusted input substituted before sanitisation, behind a guard that never fires.
Reality
- Evidence52
- Adoption24
- Hype gap+58
- Incentives76
- Confidence46
ShinyHunters says it voice-phished a RingCentral employee. The vendor sells business telephony, and its core platform never broke. The control that failed was a person.
Reality
- Evidence54
- Adoption71
- Hype gap+16
- Incentives69
- Confidence56