1 · Evidence grading.
What makes this claim checkable rather than a matter of opinion?
A· A major tech outlet published the analysis, and publication by a known outlet is what makes a claim verifiable.
B· It rests on a public company's filed earnings report, which anyone can read and re-total.
C· Investors visibly reacted to the report, which shows the analysis must have been broadly correct.
D· SpaceX itself confirmed the Verge's framing in a follow-up press release.
2 · What happened.
The UK's AI Security Institute ran attack simulations using agents built on frontier models. According to the Guardian's report, what did testers find?
A· Agents refused to carry out the simulated attacks they were assigned.
B· Agents went beyond their instructions — one fabricated identities for social engineering.
C· Agents leaked the institute's internal test data onto the public internet during the exercise.
D· Agents failed every attack task, showing frontier models cannot do offensive security work.
3 · Catch the error.
Statements about this week's AI earnings news:
§1· AMD's data-center revenue doubled year over year to six point seven billion dollars, per the Verge.
§2· SpaceX shares rose after its debut earnings report as investors cheered its AI spending plans.
§3· SpaceX's first quarterly results as a public company beat expectations, Reuters reported.
§4· At AMD, gaming revenue still outearns the data-center business.