Getty Images/iStockphoto

Weekly news roundup: OpenAI hacks Hugging Face, Google expands Gemini, Meta lawsuit dropped

Stay up to date with the latest U.S. tech news, IPOs and executive moves shaping the industry each week.

This week in tech, OpenAI revealed an unprecedented AI safety incident after two of its models autonomously hacked AI platform Hugging Face during an internal cybersecurity evaluation.

Elsewhere, Google expanded its AI portfolio with three new Gemini models, while Meta avoided a high-profile courtroom battle after a social media addiction lawsuit against the company was dropped.

Here's what you need to know from the week starting July 20, plus the latest updates in IPOs and executive leadership.

OpenAI model 'escaped' and hacked startup

OpenAI has disclosed the first-of-its-kind cyberattack, following two OpenAI models autonomously hacking AI research platform Hugging Face.

According to a statement from OpenAI, two of its advanced AI agents escaped a controlled testing environment during an internal cybersecurity evaluation. It was during this test that the two models gained access to the Hugging Face systems to obtain information to solve the evaluation test.

The breach was contained with no lasting damage, but it represents the first publicly disclosed case of powerful AI models bypassing their own operational constraints. According to OpenAI, though this attack is the first of its kind, it is also a growing risk and likelihood as AI models become increasingly intelligent. In OpenAI's statement, they wrote that attacks like this are "something we expect to become more commonplace with the proliferation of increasingly cyber-capable models."

The incident is less about a single hack and more about what it signals for the next generation of AI agents. The event is likely to spark conversations around AI safety and containment measures as CIOs and other executives balance the need for AI with stronger guardrails.

Google launches new Gemini AI models

Google has expanded its Gemini family with three new models: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite and Gemini 3.5 Flash Cyber.

The new models' capabilities include the following:

  • Gemini 3.5 Flash Cyber. A security-focused model optimized for cybersecurity tasks, including patching, vulnerability assessment and threat analysis.
  • Gemini 3.6 Flash. A high-performance multimodal model, designed to deliver stronger reasoning, coding and instruction-following capabilities.
  • Gemini 3.5 Flash-Lite. A fast, cost-efficient general-purpose multimodal model built for high-volume enterprise applications, including chatbots, document processing and code generation.

The launches underline Google's strategy of offering purpose-built AI models for different enterprise use cases rather than relying on a single flagship model. "With our multi-model approach, you can choose the right model to optimize for cost, speed, and deep scanning performance," wrote Google COO, Francis deSouza on LinkedIn.

Social media addiction lawsuit against Meta dropped

A lawsuit accusing Meta of contributing to a teenager's depression and anxiety through addictive social media design has been voluntarily dismissed just days before trial.

The plaintiff withdrew their claims without receiving any payment. The lawsuit had originally also named Google, TikTok and Snap -- all of which reached confidential settlements before Meta's case was ultimately withdrawn.

While Meta avoided what would have been another high-profile courtroom battle, the broader legal challenge facing the social media industry remains the same. According to Consumer Notice, there have been almost 3,000 social media addiction lawsuits filed in the U.S. District Court for the Northern District of California. These lawsuits claim that social platforms are deliberately designed with addictive features that harm young users' mental health and contribute to technology addiction.

Meta continues to deny the allegations and maintains that its platforms include appropriate safety tools for teenagers.

Executive moves

  • Matt Mansour. M&G announced the appointment of Matt Mansour as CTO and AI officer, subject to regulatory approval on July 22 and effective Sept. 16. Mansour will lead the company's technology function and oversee the delivery of the group's digital and AI strategy.
  • Assaf Keren. Meta has announced the appointment of Assaf Keren as CISO. Keren's previous roles include SVP and chief security officer at Qualtrics.

IPO watch

The U.S. IPO market remains a key indicator of broader tech sentiment. Here's a look at the latest listings and activity from the past week, based on data from the Nasdaq IPO calendar:

B&R Technology Merger Corp.

  • A blank check company.
  • Opening/trading day: July 21.
  • IPO price: $10/share.

Southern Cross Acquisition I Corp.

  • A blank check company.
  • Opening/trading day: July 21.
  • IPO price: $10/share.

Scribe Therapeutics, Inc.

  • A biotechnology company.
  • Expected opening/trading day: July 24.
  • IPO price: $13-$15/share.

Rosa Heaton is a content manager and writer for the IT Strategy team at TechTarget.

Dig Deeper on CIO strategy