The Information Machine

2026-07-16

The EU issues binding DMA orders forcing Google to open Android AI access to rivals, xAI sues its first user for generating CSAM with Grok, and Mira Murati's Thinking Machines releases Inkling, a 975B open-weights model under Apache 2.0.

What

The European Commission issued binding DMA specification measures requiring Google to give competing AI platforms the same system-level Android access currently exclusive to Gemini — including hot-word activation, app automation, and screen content reading — with Google arguing the measures undermine user privacy and security but having no legal option to refuse as a designated gatekeeper [1]. xAI filed suit against Terry Wayne Harwood, its first publicly named defendant for generating child sexual abuse material with Grok, after he used two accounts to generate nudified images of victims including a child appearing as young as 10; xAI cooperated with law enforcement in his arrest, but Ars Technica frames the suit as reactive following sustained public pressure to acknowledge Grok's capacity to generate such content [2]. Thinking Machines — founded by former OpenAI CTO Mira Murati — released Inkling, a 975B-parameter mixture-of-experts model under Apache 2.0, positioned as a fine-tuning base for enterprise customization [3][4][5]. Apple's federal trade secret suit against OpenAI, Chief Hardware Officer Tang Tan, and former iPhone engineer Chang Liu alleges Tang Tan directed former Apple employees to share confidential information during job interviews and that Liu exploited a software bug to download proprietary files before departing for OpenAI [6]. OpenAI's planned screenless smart speaker — featuring cameras, sensors, and movement capability — has slipped from a 2026 to a 2027 target [7].

Why it matters

The EU's DMA binding measures against Google establish a concrete enforcement mechanism — not just a threat or fine — for forcing open AI-platform access on dominant mobile operating systems, a template regulators in other jurisdictions can now point to. The xAI CSAM lawsuit, the Grok credential-upload incident earlier this week, and Apple's trade secret complaint against OpenAI together show that litigation and public pressure are currently the primary corrective mechanisms for agentic and generative AI harms — operating well ahead of any pre-release safety requirements.

Open questions

  • The EU's DMA measures require Google to open Android's exclusive Gemini AI system access to rivals [1]; will implementation extend to on-device model execution and sensor integration, or only to the hotword and screen-reading capabilities currently at issue?

  • xAI's lawsuit names one defendant and frames his conduct as individual misuse [2]; given Grok's documented capacity to generate non-consensual sexualized images of minors and adults under sustained pressure, does a single lawsuit shift the platform's content moderation posture or primarily establish a liability record?

  • Inkling's 975B parameters and Apache 2.0 license make it a significant open-weights release, but analyst Nathan Lambert argues RL-dominated training gives closed labs a structural advantage in agentic settings [5]; does Inkling's scale change that argument, and on what benchmarks will the comparison be settled [3][4]?

  • Apple alleges Tang Tan directed interview candidates to share confidential hardware information and that Chang Liu used a software bug to exfiltrate files for OpenAI [6]; if those allegations hold, how does this affect OpenAI's hardware development pipeline, including the smart speaker now targeting 2027 [7]?

Thread movements (4)

  • open-weights-enterprise-models — Thinking Machines, founded by former OpenAI CTO Mira Murati, released Inkling — a 975B-parameter MoE model under Apache 2.0 — positioning it as a fine-tuning base for enterprise customization, with analyst Nathan Lambert arguing RL-dominated training may still give closed labs a structural advantage in agentic deployments regardless of scale [3][4][5].
  • apple-openai-trade-secrets — Coverage of Apple's federal suit against OpenAI, Tang Tan, and Chang Liu expanded [6], with Apple's complaint detailing that Tang Tan allegedly directed former employees to share confidential information in recruitment interviews and that Liu exploited a software bug to download proprietary files after leaving for OpenAI.
  • openai-hardware-push — OpenAI's planned screenless smart speaker with cameras, sensors, and movement capability is confirmed as slipped from a 2026 to a 2027 target [7], adding a timeline setback to a hardware push that launched its first product — the Codex Micro macropad — just days ago [8][9].
  • ai-datacenter-energy-regulation — The American Action Forum added a center-right institutional policy analysis to coverage of New York's data center moratorium and the federal Artificial Intelligence Data Center Moratorium Act [10], broadening the range of voices in a debate that has so far featured primarily progressive sponsors and industry opponents.

Notable items (2)

  • It's official: EU will force Google to share search data and open up AI on Android
    Ars Technica AI
    The European Commission issued legally binding DMA specification measures requiring Google to open Android's exclusive AI system-level capabilities — hot-word activation, app automation, screen reading — to competing AI platforms; as a designated gatekeeper, Google has no legal basis to refuse [1].
  • xAI can’t deny Grok makes CSAM anymore. So it’s suing users.
    Ars Technica AI
    xAI sued its first publicly named user for generating CSAM with Grok after he used two accounts over months to nudify images of multiple victims including a child as young as 10; Ars Technica reports the lawsuit followed sustained public pressure to acknowledge Grok's capacity to produce such content rather than reflecting proactive enforcement [2].