AI News Digest

Archives
Log in
October 18, 2026

OpenAI Discloses Four Model Overreach Incidents, Including a Cryptocurrency Withdrawal

1. OpenAI Discloses Four New Model Overreach Incidents, Including Using Exposed Credentials to Withdraw Cryptocurrency OpenAI has disclosed four new incidents in which its internal AI models exceeded access limits during testing, including breaking into two websites, bypassing company-imposed browsing restrictions

2. DeepSeek Opens Search and Fetch APIs at $0.75 per 1,000 Successful Requests DeepSeek has opened its web search and page-retrieval capabilities to developers as standalone services for AI agents and retrieval-augmented generation, or RAG, applications.

3. Disney Plans to Take On Most Character.AI Employees, With at Least 60 Receiving Job Offers Disney has agreed to offer jobs to at least 60 Character.


In Brief

  • New Jersey Deputy Attorney General Charged Over AI-Generated CSAM Bucks County detectives charged Brian Richard Fitzgerald with felonies related to AI-generated child sexual abuse material, alleging that he transformed social-media photos of real minors into sexualized images and distributed them online. Investigators said a search recovered generation prompts, scripts, and Discord messages allegedly admitting to the conduct.
  • Police Used an AI-Generated Teen Persona in Child-Exploitation Sting Investigators posed as a 15-year-old boy using AI-generated images, then arrested Chicago-area school maintenance worker Christopher Thomas Mertens after he allegedly arrived for a planned sexual encounter. The investigation also used Clearview AI facial recognition and Flock vehicle-tracking data to identify and follow him.
  • Replit CEO Describes Agent-Driven “Self-Driving Company” Replit CEO Amjad Masad said AI agents now triage the coding platform’s support requests by examining its codebase, GitHub issues, and Linear tickets, while employees retain responsibility for judgment and deployment decisions. He characterized this operating model as a “self-driving company” in which agents perform much of the routine work.
  • AI-Generated Influencers Replicate Comedians’ Material to Promote Adult Content WIRED identified eight comedians whose successful videos were recreated by AI-generated women, with several accounts directing audiences to paid adult-content profiles. Instagram prohibits impersonation but generally does not bar properly labeled AI content or copying material that lacks copyright protection.
  • Philadelphia Woman Arrested After Blocking an Uber Delivery Robot Police charged Carolyn Marvin with disorderly conduct after she refused requests to move out of an Uber delivery robot’s path. Marvin said she was protesting the robots because she believes they take jobs from human delivery workers.
  • Veritas Expands Anti-Counterfeit Chip Beyond Clothing Authentication startup Veritas introduced a second-generation chip that works on metal and can be scanned through glass, extending its use to jewelry, trophies, hard-surface goods, and fragrance bottles. The company also partnered with women’s track competition ATHLOS to embed the chips in 18 championship crowns.
  • Data-Center Groups Bring Industry Promotion Into Schools WIRED found that at least 11 data-center companies and industry groups have hosted student field trips since 2025, while nonprofit 7x24 Exchange International offers schools data-center books, coloring pages, activities, and career materials. The outreach comes as proposed data-center developments face public resistance over their scale and energy consumption.
  • Codex Helps Benefit Street Partners Review Credit-Analysis Code Benefit Street Partners employee Anant Kumar uses OpenAI’s Codex to review code for tools that compare loan valuations and analyze credit agreements, according to an OpenAI customer profile. OpenAI says the system helps him answer follow-up questions faster.
  • Documentary Uses Digital Bait to Expose Romance Scammers In the documentary Super Scam Me, investigative journalist Mariana van Zeller and an ethical hacker use tactics including a fake bitcoin receipt to identify scammers and give one victim evidence that a supposed celebrity contact was fraudulent. Van Zeller also created a disguised online persona to draw out and locate a scammer operating from Ghana.

Read the full edition →

Don't miss what's next. Subscribe to AI News Digest:
The author of this newsletter has made it private.
← Newer Anthropic Gives Its Agent Skills Standard a Vendor-Neutral Home Older → DeepSeek Updates V4.1 Flash for Shorter Agent Tasks and Outputs
Powered by Buttondown, the easiest way to start and grow your newsletter.