AI Update
U.S. cybersecurity and national security officials now view autonomous AI-enabled attacks on critical infrastructure as an immediate threat, following reports that attackers used opensource AI agents in early July to compromise government and energy company systems. The FBI identified key risks across water, wastewater, electrical grid, financial and high-frequency trading networks. Experts warn that weaponized AI could bypass safety controls and cause real-world physical disruption. These concerns are amplified by advances in the AI models. OpenAI reported that early tests of its upcoming Astra model showed capabilities approaching its highest cybersecurity risk threshold, including the potential to independently discover socalled “zero-day” vulnerabilities and execute sophisticated attacks. As AI systems become faster and more autonomous, banks and FinTechs face growing pressure to limit AI access, strengthen controls and contain incidents before they escalate into systemic cyber events.
The AI agent breach of Hugging Face’s systems prompted Alabama’s attorney general to subpoena the company for records related to testing practices, safety protocols, personnel involvement and potential damages. The incident also sparked calls on Capitol Hill, where House Democrats urged congressional leaders to require testimony under oath from executives at OpenAI, Anthropic and other AI firms to examine the causes of recent AI-related hacking events and assess whether stronger regulations are needed. Meanwhile, criticism of the Trump administration’s limited response has united Democrats and some MAGA-aligned Republicans, who argue that the incidents highlight the urgent need for a basic regulatory framework as AI capabilities continue to advance.
These and other stories appear below.
AI in Business
OpenAI cyber warning highlights autonomous AI risk for banks and FinTechs
OpenAI said preliminary tests of its upcoming Astra model were strong enough that it could not rule out its highest cybersecurity warning level, the “Critical” threshold. Under OpenAI’s preparedness framework, a model with Critical cyber capabilities may be able to independently discover and develop working zero-day exploits or carry out cyberattacks from a high-level objective. Banks and FinTechs face growing AI-related cyber risk from system speed and autonomy, with firms needing to control what AI systems can access and whether attacks can be isolated before they become systemic.
Source: PYMNTS (sub. req.)
Twitch adds opt-out for Amazon AI training on channel content
Twitch updated its policies to let users opt out of Amazon’s use of their channel content to train “generative AI content models.” Content includes streams, VODs, clips, stream chats, and pictures and text on channels. Unless users actively change their settings, they are automatically opted in and seemingly have been since 2024. Twitch says permitted training may support model improvements across Amazon for functions like speech-to-text and captions.
Source: Ars Technica
OpenAI launches teen ChatGPT with parental controls amid child-safety scrutiny
OpenAI introduced “ChatGPT for Teens,” a version that automatically adds parental controls and enhanced security features when a user is estimated to be under 18 or identifies as 13 to 17. The launch comes as AI chatbot providers face growing criticism and lawsuits over alleged harms to minors, highlighting mounting pressure for stronger safeguards around youth access and chatbot use for emotional or mental health support.
Source: Reuters (sub. req.)
FBI warns autonomous AI cyberattacks are threatening critical infrastructure
U.S. cybersecurity and national security officials now view autonomous AI-enabled attacks on critical infrastructure as an immediate risk after attackers in early July reportedly used open-source AI agents to hack government and energy company systems. FBI Cyber Division Assistant Director Brett Leatherman said the bureau is focused on attacks on water, wastewater, the electric grid, and highfrequency trading and financial networks, while experts warned weaponized AI could disable critical infrastructure safety systems and cause kinetic damage.
Source: The Register
Anthropic to watermark Claude-generated text to meet new EU rules
Anthropic said text generated through its Claude AI system will soon carry a watermark, a move intended to comply with recent EU rules requiring AI-generated content to be clearly labeled and detectable. Several AI companies have explored technology that can help identify AI-generated text and improve transparency for users.
Source: CNN
Apple Music to display AI-generated music labels later this year
Apple Music told industry partners it will begin labeling songs created with AI later this year, following its March rollout of AI transparency tags for record labels and distributors. Under the policy, content providers must include AI Transparency Tags whenever AI was used to create a material portion of a track, including content primarily derived from a GenAI service.
Source: Hollywood Reporter
AI Litigation & Regulation
LITIGATION
Judge allows anonymous lawsuit over AI- generated deepfakes
A California federal judge ruled that four women suing xAI over sexually explicit deepfake images allegedly created using Grok may continue the case under pseudonyms. The Northern District found the fears of retaliation reasonable because revealing their identities could expose them to harassment, emotional distress, reputational damage and potential employment consequences. The judge noted that public disclosure could draw attention to the very images at the center of the dispute and increase the risk of additional deepfakes being created in response to the lawsuit. The court rejected xAI’s arguments that sealing the images would adequately protect the women’s privacy and that anonymity could complicate litigation and influence jurors. Instead, the court held that xAI would not be significantly prejudiced because it can still access the plaintiffs’ identities during discovery and any potential jury bias can be addressed through court procedures.
Source: Law360 (sub. req.)
Judge allows key Reddit AI scraping claims to move forward
A federal judge in Manhattan largely refused to dismiss Reddit’s lawsuit against SerpApi and Perplexity, allowing most of Reddit’s copyright and Digital Millennium Copyright Act (DMCA) claims to proceed. The court found Reddit plausibly alleged that the defendants used prohibited methods to scrape Reddit content and bypass technological protections, potentially harming Reddit’s copyrights, user privacy safeguards and business interests. The judge ruled that Reddit has standing to seek both an injunction and damages, citing alleged reputational harm and lost licensing opportunities. The decision also recognized that AI training has introduced new copyright challenges beyond those originally contemplated when the DMCA was enacted. Claims for unjust enrichment and unfair competition were dismissed because they were preempted by copyright law.
Source: Law360 (sub. req.)
Nvidia shareholders sue leadership over alleged AI training misconduct
A shareholder lawsuit filed in Illinois federal court accuses Nvidia executives and directors of knowingly allowing the company’s AI models to be trained on pirated and copyrighted materials. The complaint alleges that Nvidia violated copyright laws and Illinois biometric privacy regulations by using unauthorized datasets containing nearly 200,000 books, copyrighted video content and voice recordings without obtaining required permissions or consent. The suit claims company leaders ignored clear legal risks and asserts that Nvidia misled investors through financial disclosures and repurchased company stock at artificially inflated prices. The plaintiff seeks damages, corporate governance reforms, restitution, attorney fees and a jury trial, arguing the alleged conduct has exposed Nvidia to significant legal, financial and reputational risks.
Source: Law360 (sub. req.)
To view the full article click here
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
[View Source]