Musk's Shocking Revelations on Grok

Summary of Musk's Shocking Revelations on Grok

by Candace Fan

12m•May 1, 2026
Listen to Episode

Overview of Musk's Shocking Revelations on Grok

This episode is a fast-moving update on the AI industry’s biggest power plays: Google’s Gemini expanding into millions of GM and Volvo vehicles, Anthropic’s new Claude Security beta and its tense relationship with the Pentagon, Elon Musk’s courtroom admission that xAI partly used OpenAI outputs to train Grok, and the U.S. government’s latest list of AI vendors approved for classified networks. The throughline is clear: AI competition is now happening across consumer products, enterprise security, and national defense at the same time.

Major Topics Covered

Gemini expanding into GM and Volvo cars

  • Google announced that Gemini is replacing Google Assistant in about 4 million GM vehicles via Android Auto / Google built-in.
  • Volvo was also added, with the rollout covering 16 models going back to 2020.
  • The upgrade is over the air, so owners do not need to replace hardware.
  • The host emphasized the practical appeal: better in-car voice control, especially with Google’s stronger integration across Maps, Gmail, and Docs.

Anthropic launches Claude Security in public beta

  • Claude Security is now in public beta for Claude Enterprise users, with broader rollout planned for Teams and Max.
  • The tool scans codebases like a security researcher would:
    • traces data flows,
    • identifies how components interact,
    • flags vulnerabilities,
    • recommends fixes,
    • and provides remediation steps.
  • It can connect to a GitHub repo, generate a report, and hand off actionable output to Claude Code for implementation.
  • The feature also supports:
    • scheduled scans
    • dismissal of low-priority findings
    • CSV and markdown exports
    • routing results into ticketing systems
  • Anthropic’s ecosystem partners mentioned include:
    • CrowdStrike
    • Palo Alto Networks
    • SentinelOne
    • Trend Micro
    • Wiz
    • plus consulting firms like Accenture, BCG, Deloitte, Infosys, and PwC

Elon Musk’s courtroom admission about Grok training

  • In federal court, Musk was asked whether xAI used OpenAI model outputs to train Grok.
  • He acknowledged that xAI did partly use distillation from OpenAI models.
  • The host framed this as common industry practice, though it is against OpenAI’s terms of service.
  • The episode notes that this practice is widespread in AI training, even if it is controversial.

Pentagon names seven AI companies for classified network access

  • The Pentagon reportedly named seven companies it will allow into its classified network:
    • OpenAI
    • Google
    • NVIDIA
    • SpaceX
    • Reflection AI
    • Microsoft
    • AWS
  • These companies are said to be operating in impact levels 6 and 7, the highest classification tiers tied to secret and top-secret work.
  • Anthropic was notably excluded, which the host sees as the bigger story.
  • The exclusion follows a dispute in which the Pentagon had labeled Anthropic a “supply chain risk” after the company resisted unrestricted government access.
  • A federal judge had already blocked that designation earlier in the year.

Anthropic’s funding timing and market position

  • Anthropic is reportedly pushing a major funding round at a $900 billion valuation and has given investors 48 hours to confirm participation.
  • The host interprets this as a sign that Anthropic is nearing a major inflection point, possibly its last private round before an IPO.
  • The episode suggests the company is in high demand, despite the Pentagon tension.

Reflection AI as a strategic substitute

  • Reflection AI was highlighted as an interesting name on the Pentagon’s approved list.
  • Its founders come from DeepMind, with backgrounds in Gemini reward modeling and AlphaGo architecture.
  • The host argues that including Reflection alongside OpenAI, Microsoft, and other giants may signal that the Pentagon is willing to support frontier-model alternatives if needed.
  • In that framing, Reflection serves as a kind of stand-in for Anthropic after the exclusion.

Key Takeaways

  • AI is becoming embedded in everyday products: cars, code editors, and enterprise workflows are all being upgraded with advanced models.
  • Security is a major battleground for AI companies, and Anthropic is positioning Claude Security as a flagship enterprise offering.
  • Training practices remain contentious: Musk’s admission reinforces how common model distillation is, even when it violates platform rules.
  • Government access is now a competitive advantage: the Pentagon’s vendor list highlights which AI firms are trusted for sensitive national-security work.
  • Anthropic appears both challenged and in demand: shut out of one major defense channel, but still highly valued by investors and enterprise buyers.

Notable Strategic Implications

For Google

  • Gemini’s vehicle rollout expands Google’s AI presence into a huge installed base of cars.
  • The integration could make the car a more useful voice-controlled productivity environment, not just a navigation tool.

For Anthropic

  • Claude Security strengthens Anthropic’s position in enterprise cybersecurity.
  • The Pentagon exclusion could be a setback, but the company is still being treated as a major player in commercial security and AI tooling.

For xAI and OpenAI

  • Musk’s admission may fuel further debate over data sourcing, model distillation, and competitive fairness.
  • It also underscores how blurred the line is between “benchmarking,” “distillation,” and model reuse in the current AI race.

For U.S. defense procurement

  • The approved vendor list suggests the Pentagon is splitting its AI partnerships across both established giants and newer frontier labs.
  • Anthropic’s absence, alongside Reflection AI’s inclusion, signals that alignment with defense priorities is now a market differentiator.