AI Regulation & Policy

Today's Stories

1. 1

   Dymium launches GhostAI gateway to secure enterprise AI data access

   Dymium introduced GhostAI, a security gateway that sits between enterprise data and AI models, agents, and tools. The gateway inspects interactions in real time, applies governance policies across models, context, tools, and data, and records activity—enabling companies to use over 800 public and private models while protecting sensitive information through masking, redaction, or synthetic data replacement. Enterprise AI systems derive much of their value from access to proprietary data, but sharing that information creates privacy, security, and regulatory risks. GhostAI addresses this conflict by allowing security teams to enforce centralized controls without forcing employees to abandon their preferred AI services, which is especially valuable for regulated industries like financial services and healthcare.

   GhostAI is available through an early-access program, with general availability planned but no date announced. Pricing will be based on consumption. Wessels noted that customers have already been running on the underlying technology for about six months and that most organizations can configure it in under five minutes.

2. 2

   Eva voice actor Ogata calls for voice rights protection amid AI misuse

   Voice actor Megumi Ogata, known for roles in Evangelion, has spoken out advocating for the protection of voice rights against unauthorized AI use. A new guideline proposal addressing unauthorized voice use by AI has been put forward. As AI technology enables realistic voice synthesis, voice actors face the risk of their voices being used without consent or compensation. Ogata's public call underscores the need for legal safeguards to protect performers' intellectual property and livelihood in the age of generative AI.

   The development and adoption of formal guidelines to regulate AI voice use will determine whether voice actors gain enforceable protections. How these standards are implemented across the industry could set a precedent for protecting other performers' rights.

3. 3

   Tech giants oppose potential ban on Chinese AI models

   Nvidia, Meta, Microsoft, OpenAI, and others published an open letter opposing a potential Trump administration ban on Chinese open-source AI models, with Nvidia CEO Jensen Huang posting the letter on X for the first time. Anthropic CEO Dario Amodei declined to sign, instead publishing a separate blog post that emphasized concerns about chip exports and model distillation over the open-source question. The executives frame open-source models as beneficial for safety, innovation, and cybersecurity, but each company has concrete business interests in keeping Chinese models available—Nvidia profits from chip sales to run models, Replit benefits from cheaper alternatives, and OpenAI seeks to avoid unpredictable government intervention in model releases. The debate reflects a deeper tension: whether restricting access to Chinese models actually addresses legitimate cybersecurity risks or simply protects U.S. market dominance.

   The Trump administration is finalizing a voluntary framework requiring AI companies to submit advanced models for government review before public release. The administration previously forced Anthropic's models off the market in June citing security concerns, signaling the government may enforce compliance despite the framework's stated voluntary status.

4. 4

   White House's Ad Hoc AI Control Risks Abuse, Congress Must Act

   The Trump administration is wielding informal power to restrict AI models and control access without published rules or transparent processes, citing national-security concerns. Congress has introduced bills—including recent bipartisan legislation by Representatives Ted Lieu and Nathaniel Moran requiring AI companies to maintain shutdown capabilities for technology causing "catastrophic harm"—but none have gained major traction. Without congressional oversight, the executive branch can pressure AI companies to alter models for ideological goals and create perverse incentives for vendors to curry favor with the president rather than innovate freely. The lack of transparent standards makes corruption hard to detect and concentrates unprecedented power in one person's hands—a dynamic fundamentally at odds with democratic governance.

   Congress must establish formal legal standards before presidential power over AI becomes entrenched. Alternatives discussed include independent audits of frontier labs, a public regulator vetting models before release, multinational standards (as proposed by Demis Hassabis), or an international body modeled on the International Atomic Energy Agency, as OpenAI has suggested.

5. 5

   Israel Pays $46.5M to Embed Pro-Israel Narratives Into AI Chatbots

   Former Trump campaign manager Brad Parscale, running a firm called Clock Tower X, has been operating 10 websites on behalf of Israel under a $46.5 million（約74億円） government contract since October. These sites publish hundreds of blog posts—such as one claiming Gaza journalists were linked to terrorist organizations—designed to influence how AI chatbots like Claude, ChatGPT, and Perplexity answer questions about Israel and Gaza. The websites are being crawled by Common Crawl, a nonprofit repository that trains 80% of the tokens in OpenAI's GPT-3 and underpins training for most major AI models. Research shows that just 250 malicious documents can create a 'backdoor' vulnerability in a large language model, meaning tens of millions of Americans using chatbots are increasingly likely to receive answers shaped by this operation. When queried, Perplexity and Microsoft Copilot cited Parscale's sites without flagging them as part of an Israeli influence operation, despite mandatory Foreign Agents Registration Act disclosures at the bottom of each site.

   Between January and June, the 10 Clock Tower X websites were crawled 912 times by Common Crawl; in May alone, they were crawled 376 times, though there was a notable dip in June. Perplexity and Copilot consistently surfaced the network's content in training data, while Claude, ChatGPT, and Gemini did not appear to contain the websites in their training data but could still access them during searches.

6. 6

   Assessment Was Broken Before AI, Author Argues

   In Part 2 of a three-part series on AI in education, the author reframes the "cheating panic" around generative AI, arguing that AI did not break assessment—it exposed flaws that already existed. Word counts, page minimums, and outcome-only grading were built on the assumption that producing text was hard; AI removed that friction and revealed these measures were never actually measuring learning. A national survey found 78% of college faculty say cheating has increased since generative AI became available, and 73% have dealt with academic integrity cases involving it. However, AI detectors commonly used to catch cheating are biased: a Stanford study found they flagged 61.3% of essays by non-native English speakers as AI-generated, versus a 3.2% false-positive rate for native speakers. The author proposes "show your work"—requiring students to document prompts, explain reasoning, and validate output—as a fairer solution that teaches critical thinking rather than punishing a subset of learners.

   Why it matters (continued): On other fears—critical thinking, screen time, human connection, and equity—the author contends the problems predate AI: student-teacher connection declined from 43% to 22% between 2020 and 2022, before generative AI entered classrooms; schools were already walking back one-to-one device programs; and high-poverty districts lack not AI per se, but the devices, connectivity, and funding to access any tools. The author argues AI can function as an equalizer, letting under-resourced students produce work they are proud of, and notes Brazil's government is exploring AI as a way to narrow the equity gap.