As the standoff between the United States government and Minnesota continued this week because of the immigration enforcement operations that actually occupied the Twin Cities and other parts of the state, a federal judge delayed a decision this week and ordered a new briefing on whether the Department of Homeland Security using armed raids to pressure Minnesota to abandon sanctuary policies for immigrants.

Meanwhile, minutes after a federal immigration officer shot and killed 37-year-old Alex Pretti in Minneapolis on Saturday, Trump administration officials and right-wing influencers. A smear campaign has been launched, calling Pretti a “terrorist” and a “crazy.”

As part of his surveillance dragnet, Immigration and Customs Enforcement has been using an AI-powered Palantir system since last spring to summarize tips sent to its tip line, according to a newly released Homeland Security document. DHS immigration agents also use the now-famous facial recognition app Mobile Fortify to scan the faces of countless people in the US—including many citizens. And a New ICE filings provide insights into how commerce worksincluding ad tech and big data analysis, is increasingly being considered by the government for law enforcement and surveillance. And the Active duty military officer breaks down federal immigration enforcement actions in Minneapolis and around the US for WIREDconcluding that ICE is masquerading as a military force, but is actually using immature tactics that can kill real soldiers.

WIRED published extensive inside details this week on the inner workings of a scam compound in the Golden Triangle region of Laos after a human trafficking victim calling himself Red Bull spoke to a WIRED reporter for months and leaked a large trove of internal documents from the compound where he was being held. Importantly, also WIRED records his own experiences as a forced laborer in the compound and his attempts to escape.

Deepfake “naked” technology and tools that produce sexual deepfakes are becoming increasingly sophisticated, capable, and accessiblewhich poses an increased risk for the millions of people abused by technology. Moreover, this week’s research found it an AI stuffed animal toy from Bondu has its web console virtually unprotected, exposing 50,000 logs of children’s chats to anyone with a Gmail account.

And there are others. Each week, we round up the security and privacy news we don’t quite understand. Click on the headlines to read the full stories. And stay safe out there.

According to a document released by the Department of Justice on Friday, an informant told the FBI in 2017 that Jeffrey Epstein had a “personal hacker.” The document, first reported by TechCrunch, was released as part of a large trove of material the DOJ is legally required to release in connection with the investigation into the late sex offender. The document does not provide the identity of the alleged hacker, but it includes some details: They are said to have been born in Italy in the southern region of Calabria, and their hacking focused on discovering vulnerabilities in Apple’s iOS mobile operating system, BlackBerry devices, and the Firefox browser. The informant told the FBI that the hacker was “very good at finding vulnerabilities.”

The hacker allegedly created offensive hacking tools including exploits for unknown and/or unpatched vulnerabilities and allegedly sold them to several countries, including unnamed central governments in Africa, the UK, and the US. The FBI informant further reported that the hacker sold an exploit to Hezbollah and received “a tree of cash” in payment. It is unclear whether the informant’s account is accurate or whether the FBI has verified the report.

The viral AI assistant OpenClaw—once called Clawdbot and then, briefly, Moltbot—took Silicon Valley by storm this week. Technologists allow the assistant to take control of their digital lives: connecting it to online accounts and allowing tasks to be completed for them. the assistant, as reported by WIREDruns on a personal computer, connects with other AI models, and can be granted access to your Gmail, Amazon, and many other accounts. “I can automate anything. It’s amazing,” one entrepreneur told WIRED.

They are not the only ones intrigued by the capable AI assistant. The creators of OpenClaw TELL more than 2 million people visited the project last week. However, its agent abilities have potential security and privacy trade-offs—beginning with the need to provide access to online accounts—that may be impractical for many people to operate securely. As OpenClaw has grown in popularity, security researchers have identified “hundreds” of instances where users have exposed their systems to the web, the Registration reported. Many do not include authentication and expose full system access to users.