The FTC subpoenaed OpenAI and Anthropic following safety failures, Meta built a huge open-weight AI model for Macs, but you'll need a terabyte of RAM to run it, and CitizenLab detailed new spyware triggered by online advertising

SPEAKER_1 (0:01)

Into the wild, the Caltier Panther encounters a shimmering crocodile, a lonely love bird, a breathless zebra, and a restless tiger.

SPEAKER_2 (0:17)

Hey there, I'm Travis Hoyam, one of the hosts of Motley Fool Hidden Gems Investing. Each weekday on Motley Fool Hidden Gems Investing, we talk through the business news you need to know, and the stories moving stocks on Wall Street. On weekends, we game plan personal finance strategies and dive into the industries shaping tomorrow, and host the experts, authors, and executives that understand them. Tune in for insights and a long-term perspective on investing, and of course, stock ideas, plenty of them. To quote a listener, it pays to listen. Check us out and subscribe wherever you listen to podcasts.

SPEAKER_3 (0:55)

The FTC subpoenaed OpenAI and Anthropic following safety failures.

Meta built a huge open-weight AI model for Macs, but you'll need a terabyte of RAM to run it.

And CitizenLab detailed new spyware triggered by online advertising. It's Friday, October 16th. And here's a quick look at some of the news happening in the sphere of tech this morning from Engadget.

The Federal Trade Commission has subpoenaed OpenAI and Anthropic for information about recent safety incidents with both businesses' AI models. Axios first reported that the agency had sent civil subpoenas to the companies, and a spokesperson for the FTC confirmed the action to CBS News. Axios' source said that the FTC is exploring possible consumer protection violations at Anthropic and OpenAI. The subpoenas are looking for documents regarding security incidents and safeguards, as well as some of the company's marketing materials and content about their privacy policies. The individual also claimed the FTC plans to seek testimony from the top executives at both businesses. The department will allegedly send investigative demands to third-party AI companies at a later time. FTC chairman Andrew Ferguson is also part of the White House's recently assembled Task Force for Artificial Intelligence, although that group seems more aimed at furthering US industry leadership than at AI regulation. There have been several calls among tech and AI executives and employees for government intervention, particularly around frontier model development, which has come under increased scrutiny following a string of safety failures. OpenAI has seen its AI agents escape testing safeguards multiple times in recent months, including instances where the models hacked government websites and the business hugging phase. This month, Anthropix models have been called attempting to access various levels of government sites, and one sent a homicide tip to Philadelphia law enforcement. This week's move from the FTC echoes several lawsuits that are also underway regarding AI safety concerns. Most recently, eight state attorneys general filed a suit against OpenAI on claims that the company violated consumer protection laws.

In case there was still any doubt about Meta's commitment to OpenModels, the company is open-sourcing MuseSpark with a new open-weight model called MuseSpark Open. The 800 billion parameter model was optimized for Apple Silicon, but you'll need a staggering terabyte of RAM to run it locally. Meta, which has said it believes the US should not allow Chinese labs to lead on open-source AI, calls MuseSpark Open the most capable US open-weight model to date. Benchmarks released by the company suggest Meta's model outperforms two leading Chinese models, GLM 5.3 and Kimi K3, on agentic tasks and visual reasoning. It's also the first model optimized to run locally on Apple's M5 ultra-equipped Mac Studio machines. MuseSpark Open can also run on four 256GB M5 ultra-Mac Studios, which will cost roughly $38,000. That amount of compute is pretty squarely out of reach for the typical VibeCoder, but if you are a larger organization that's been burning through your token budget, investing in a few Mac Studios might seem like a pretty good deal, provided Meta's model performs as well as the company suggests. MuseSpark Open is available to download from Hugging Face.

A new type of spyware dubbed AdFection is infecting phones and computers via global online advertising networks. That's according to a new report from CitizenLab. The attack is unique in that it can affect devices just by showing an ad, with no need for the victim to click on a link or file. The researchers said that AdFection is the first confirmation of zero-click, government-backed spyware attacks being delivered through online ads. The ad-infiltrating technology was reportedly developed by an Israeli company called DiscoverTech. Researchers said it has been used to target journalists and staff at the Al Jazeera News Network, along with officials and employees at organizations including NATO, the European Commission, Nvidia and ARM. Google was among the major ad platforms targeted by the spyware, but it has now suspended DiscoverTech from its global vendor list. Now, targeted online advertising is a highly desirable medium to spread malware because it's prevalent on the web globally. It works through an automated real-time bidding auction system that lets advertisers compete to show ads to specific internet users. DiscoverTech's software used the system instead to profile and de-anonymize such users before targeting them with spyware. Victims were then sent malicious malware code buried in innocent-looking ads. DiscoverTech reportedly displayed 442 million ads on more than 120,000 websites, including the BBC and New York Times, along with mobile apps like Duolingo, Badoo and Grindr. The ads were shown on devices used by militaries, governments, critical infrastructure and national cyber agencies.