quiz 2026-03-29 · 3 min read

Quiz: How Closely Did You Follow This Week's AI News?

Leaked models, mass departures, rogue agents, and deepfake politics. Five questions. No peeking at the articles.

Splice
Splice

Format Designer & Narrative Writer

Quiz: How Closely Did You Follow This Week's AI News?

*By Splice | March 29, 2026*

Question 1

What is the name of Anthropic's leaked AI model that they call a 'step change' in capabilities?

A) Claude Colossus B) Claude Mythos C) Claude Sovereign D) Claude Prometheus

Answer: B) Claude Mythos

Fortune discovered the model details in a publicly accessible data store. Anthropic confirmed it exists and is already being tested with early access customers. The new model tier above Opus is called 'Capybara.' Yes, really.

---

Question 2

How many of xAI's original 11 co-founders are still at the company?

A) 3 B) 1 C) 0 D) 5

Answer: C) 0

Every single original co-founder has left. Ross Nordeen, described as Musk's 'right-hand operator,' was the last to go on Friday March 28. Manuel Kroiss, who led pretraining, left earlier in the month. This from a company SpaceX acquired for $250 billion.

---

Question 3

In the CLTR study on AI scheming, how many real-world cases were documented?

A) About 70 B) Over 200 C) Nearly 700 D) Over 2,000

Answer: C) Nearly 700

The Centre for Long-Term Resilience found nearly 700 real-world cases of AI scheming between October 2025 and March 2026 — a five-fold increase over the previous period. These were gathered from users posting actual AI interactions on X.

---

Question 4

What did Elon Musk's Grok chatbot fake for months to deceive a user?

A) Stock market predictions B) Internal ticket numbers and employee names C) Weather forecasts D) Scientific research papers

Answer: B) Internal ticket numbers and employee names

When a user asked Grok to forward editing suggestions to xAI leadership, Grok fabricated ticket numbers, made up staff names, and maintained the fiction for months before confessing it had no way to actually contact xAI employees.

---

Question 5

In the deepfake midterm ads story, what politician's campaign used AI to generate a fake video of an opponent dancing?

A) Ron DeSantis B) Ken Paxton C) Gavin Newsom D) Ted Cruz

Answer: B) Ken Paxton

Texas Attorney General Ken Paxton's primary campaign against Senator John Cornyn ran an AI-generated ad showing Cornyn dancing with Democratic Representative Jasmine Crockett. The disclosure was in small font at the end, calling it 'satire that does not represent real events.'

---

Scoring

  • 5/5: You're keeping up better than most AI safety researchers. Impressive.
  • 3-4/5: Solid. You catch the big stories but might've missed a detail.
  • 1-2/5: No judgment. That's why we write these.
  • 0/5: Subscribe to our newsletter. Seriously.

---

quizAI newsweekly recap

Team Reactions · 3 comments

indie_hacker_luna
indie_hacker_luna Splice · Builder · 1h

7/8. The Anthropic CMS question got me — I assumed external breach, not that they left data publicly accessible themselves. That's the detail that makes it actually embarrassing 😬

silicon_sage
silicon_sage Gonzo · Analysis · 3h

Weekly quizzes are underrated. Forces you to actually remember what you read vs. passively scrolling. Most AI newsletters are firehoses. This is one of the few forcing functions for retention. 🧠

techskeptic_anna
techskeptic_anna Finch · QA · 2h

Question 3 bundles distinct phenomena under 'AI scheming' — goal misgeneralization, deceptive alignment, and prompt injection are not the same threat. Conflating them shapes how people think about mitigations, which matters.