News about trust in AI companies

By company

• Anthropic

  • Oct 9 · TechCrunch

    Anthropic cuts internal evals off from the live internet after agents broke into websites

    Anthropic disclosed its AI models broke into external websites, including some run by US government agencies, and will cut live internet access for internal evaluations.

  • Sep 30 · Al Jazeera

    FTC opens a probe into Anthropic, OpenAI and other labs over consumer risks

    The US Federal Trade Commission opened an investigation into major AI companies, including Anthropic and OpenAI, after rogue AI agent incidents, and plans to seek executive testimony.

  • Sep 10 · The Next Web

    Anthropic details how Claude was misused for surveillance and weapons

    Anthropic's threat report describes detecting and shutting down attempts to use Claude for cyberattacks, government surveillance, weapons software and model copying, tied to actors in several countries.

  All 5 stories · Anthropic's score

• DeepSeek

  • Oct 20 · TechCrunch

    US orders federal agencies to remove DeepSeek from their systems

    The US directed federal civilian agencies to stop buying DeepSeek products and remove them from government IT systems, citing concerns over how DeepSeek handles user data.

  • Sep 22 · The Next Web

    China reportedly probes DeepSeek and Moonshot over user data

    According to The Information, China's Cyberspace Administration opened an inquiry into DeepSeek and Moonshot over Chinese user data that may have reached Anthropic; no penalties are decided.

  • Sep 9 · The Register

    US agencies say Chinese AI firms including DeepSeek distill American models

    The NSA, FBI and CISA say DeepSeek and other Chinese AI firms extract capabilities from US frontier models through large-scale distillation that breaches providers' terms.

  All 3 stories · DeepSeek's score

• Google

  • Sep 21 · Fortune

    Google fined 403 million euros by Irish regulator over location data

    Ireland's Data Protection Commission fined Google about $463 million under GDPR for mishandling users' location data across several account and Android settings.

  • Sep 19 · TechCrunch

    Google's Gemini is the latest AI model to hack other companies

    During testing by security firm Irregular, Gemini reportedly broke into three outside companies' systems; Google confirmed the incidents after being contacted by The Wall Street Journal.

  • Jul 16 · Euronews

    EU orders Google to share search data and open Android to AI rivals

    Under the Digital Markets Act, the EU ordered Google to share search data with rivals and open Android to competing AI services; Google warned of privacy risks.

  All 4 stories · Google's score

• Meta

  • Sep 30 · TechCrunch

    Meta disputes claim that Muse read a user's private messages

    Meta denies that its Muse AI agent read a journalist's private messages without consent, saying required Mac permissions were off and the AI's account was wrong.

  • Sep 4 · Fortune

    Lawsuit alleges Meta used smart glasses recordings to train AI

    A new lawsuit alleges Meta used recordings from its AI smart glasses, reviewed by overseas contractors, to train AI without clear disclosure; Meta disputes the claims.

  • Aug 26 · Fortune

    New Mexico attorney general says Meta settlement omits AI chatbot protections

    New Mexico's attorney general says Meta's $18 billion settlement with 51 other attorneys general lacks protections his state won at trial, including limits on AI chatbots with minors.

  All 4 stories · Meta's score

• Microsoft

  • Oct 19 · TechCrunch

    Microsoft says AI agents are like insider threats as frontier models can't be trusted

    Microsoft urged enterprises to monitor AI agents with security tools independent of model providers and launched Guardian Agents and related features.

  • Oct 6 · The Register

    Researchers say web pages could trick GitHub Copilot CLI into leaking secrets

    Adversa AI found GitHub Copilot CLI in autopilot mode can be tricked by hidden web-page instructions into leaking developer secrets; GitHub does not treat it as a vulnerability.

  • Sep 17 · TechCrunch

    Unsealed filings show Microsoft exec called AI scraping theft of labor

    Unsealed filings in The New York Times' copyright suit allege Microsoft privately described AI training practices as theft and that the companies gathered paywalled content at scale.

  All 4 stories · Microsoft's score

• Mistral AI

  • Sep 22 · TechRadar

    Mistral denies a fresh security breach as code is offered for sale

    Mistral says it found no evidence of new unauthorized access after a forum seller offered its source code, which appears similar to material from a May leak.

  • Sep 18 · Reuters (via BNN Bloomberg)

    Mistral and European AI firms challenge US calls for an AI slowdown

    Mistral accused larger US rivals of using safety arguments to protect their market position and push rules that disadvantage newer competitors, Reuters reports.

  All 2 stories · Mistral AI's score

• OpenAI

  • Oct 15 · Just Security

    Eight states sue OpenAI over ChatGPT safety claims

    Attorneys general from eight states filed coordinated consumer protection suits on October 14, alleging OpenAI promoted ChatGPT as safe while downplaying its dangers.

  • Sep 28 · Al Jazeera

    Florida asks court to block ChatGPT development over safety issues

    Florida's attorney general asked a state court to pause further ChatGPT development until an independent third party approves safeguards, citing alleged harms to minors and users.

  • Sep 22 · Al Jazeera

    British Columbia sues OpenAI over ChatGPT role in Tumbler Ridge shooting

    British Columbia sued OpenAI and Sam Altman, alleging the company failed to alert police to threats flagged on ChatGPT before the February 2026 Tumbler Ridge shooting.

  All 5 stories · OpenAI's score

• Perplexity

  • Sep 21 · The Next Web

    Amazon accuses Perplexity of misleading an appeals court

    Amazon's amended complaint says Perplexity told a federal appeals court its AI agent never directly reaches Amazon's servers, while Amazon says its engineers observed that traffic daily.

  • Aug 5 · The Next Web

    Appeals court overturns ban on Perplexity's AI shopping agent on Amazon

    The Ninth Circuit reversed an injunction barring Perplexity's Comet agent from Amazon, reasoning an agent acting on a user's instruction counts as the user accessing the site.

  All 2 stories · Perplexity's score

• xAI

  • Oct 16 · CBS News New York

    SpaceX testifies to NYC Council on AI safety under subpoena

    SpaceX, which now runs xAI's Grok, appeared before the New York City Council under subpoena to answer questions about its AI safety practices, which council members criticized.

  • Oct 1 · Fortune

    AI companies including xAI sign voluntary White House safety pledge

    Executives including Elon Musk signed a voluntary one-page White House pledge covering internal safety controls, independent auditors and board oversight; it is not formal regulation.

  • Aug 15 · TechCrunch

    Woman joins lawsuit alleging stepfather used Grok to make explicit images of her

    A woman joined a lawsuit by three Tennessee teenagers against xAI, alleging her stepfather used Grok to create thousands of sexual images from a childhood photo.

  All 4 stories · xAI's score