Tech Grapples With ICE + Casey Tries Clawdbot, a Risky New A.I. Assistant + HatGPT

What It Covers

Hard Fork examines tech's response to ICE operations in Minneapolis, including CEO statements, AI-manipulated images from the White House, and surveillance infrastructure. Casey Newton tests Multbot, an open-source AI agent that controls computers locally but poses security risks. The hosts discuss AI-generated misinformation, platform responsibility, and the widening gap between early AI adopters and cautious users.

Key Questions Answered

• •Tech CEO Political Positioning: Sam Altman, Dario Amodei, and Tim Cook issued carefully worded statements about Minneapolis ICE operations to employees, balancing workforce pressure against White House retaliation risk. Altman called ICE actions overreach in internal Slack, while Cook urged de-escalation. These minimal statements reflect CEOs operating as quasi-politicians managing hundreds of millions of users while avoiding antagonizing the administration, which could trigger regulatory or public attacks.
• •Government Content Production Strategy: ICE now employs dedicated content creation teams using paid social media tools, producers, editors, and video makers to control narratives around enforcement operations. Internal documents reveal the agency treats winning on social media as a primary objective, not secondary to policy goals. This represents a shift where federal agencies function like brands or celebrities, actively shaping public perception through coordinated digital campaigns rather than simply enforcing law.
• •AI Image Manipulation in Politics: The White House released AI-altered images during Minneapolis protests, including a doctored photo making civil rights attorney Nekima Levy Armstrong appear to be crying when she was not. Vice President JD Vance retweeted the manipulated image. When questioned, a White House spokesman stated the memes will continue, demonstrating deliberate use of fabricated evidence to shape public opinion and erode trust in visual documentation of events.
• •Multbot Security Trade-offs: Multbot is an open-source AI agent running locally on computers with access to email, calendars, and system controls. Users face prompt injection attacks where hidden website instructions trigger unwanted actions, and messaging app integrations create entry points for complete computer takeover. San Francisco tech workers bought separate Mac Minis to sandbox the tool, recognizing the risk of granting AI agents unrestricted access to personal data and accounts.
• •Smartphone Documentation Dynamics: Minnesota Governor Tim Walz urged citizens to carry phones constantly to document federal agent actions, while DHS Secretary Kristi Noem declared videotaping ICE operations constitutes doxxing and threatened prosecution. This creates phone-versus-phone confrontations where protesters document potential abuses while the administration brings conservative influencers to produce counter-narratives. The administration appears more concerned about documentation than physical threats, with agents remaining permanently masked during operations.
• •AI Adoption Polarization Risk: A growing divide separates early adopters installing experimental AI tools like Multbot from organizations still debating basic chatbot policies. Legendary AI researcher Andrej Karpathy states tools like Claude Code represent the biggest change to his programming workflow in twenty years, occurring within weeks. This gap may create competitive advantages for those embracing AI coding assistants while others continue manual work, though productivity gains remain debatable outside specialized domains like software development.