DHS Orders Federal Agencies to Ban Chinese AI Startup DeepSeek Over Security Concerns – DTH

U.S. Orders Federal Agencies to Remove DeepSeek AI

The U.S. Department of Homeland Security has ordered federal civilian agencies to purge and cease purchasing products or services from Chinese AI startup DeepSeek due to national security and data privacy concerns. This order, expanding on previous intelligence community restrictions, also affects third-party systems built on DeepSeek’s low-cost open-weight models and follows allegations of unauthorized model distillation from U.S. rivals like OpenAI and Anthropic. In response, China’s Foreign Ministry criticized the U.S. for politicizing technology, while Beijing reportedly considers its own export restrictions on advanced AI tools.

Source: TechCrunch

Qualcomm Plans Worldwide Job Cuts

Qualcomm plans to reduce its global workforce following annual performance reviews to foster a stronger performance culture and align skills with evolving business priorities. The chipmaker has faced slumping consumer electronics sales and slower share growth in 2026, alongside forecast shortfalls in Apple revenue. However, Qualcomm expects non-handset businesses, particularly in the expanding AI data-center market, to offset these losses and drive the majority of chip sales by fiscal 2027.

Source: Reuters

Meta Revises Website Scraping Terms After Backlash

Meta updated its commercial terms of service after facing backlash from publishers and legal experts over language granting broad permission to scrape business account websites. Although Meta removed explicit mention of AI “training” and stated the changes only apply to advertising and business products, critics argue the updated terms still significantly expand Meta’s scraping rights and spark serious legal and privacy concerns across the media industry.

Source: Engadget

Apple Commits $100 Million to U.S. Titanium Manufacturing

Apple has committed to purchasing up to $100 million of recycled titanium over three years from Pennsylvania-based Perryman Company to support products like the foldable iPhone Duo. This investment is part of Apple’s American Manufacturing Program, which also includes expanding partnerships with Qnity Electronics for camera system films, sourcing chips from TSMC’s Arizona plant, and manufacturing cover glass at Corning’s Kentucky facility.

Source: Reuters

OpenCode Security Breach Exposes Users’ API Keys

Open-source AI coding agent, OpenCode, experienced a security breach affecting 1.4% of users on its online Console platform, resulting in exposed API keys and unauthorized charges across various AI services. During the incident, the team faced challenges defending against the attack as commercial AI safety filters blocked their analysis of malicious payloads. Although access to OpenAI’s Daybreak Blue program eventually aided the investigation and key providers moved to revoke compromised credentials, the incident underscores the growing complications and security risks associated with AI-driven cyber threats.

Source: TechCrunch

OpenAI Targets Legacy Code Modernization With AI

OpenAI is enhancing its AI models to assist enterprise customers and banks with modernizing legacy systems written in older programming languages like COBOL, Fortran, and Visual Basic. While models like GPT-6 Astra and GPT-6.1 Sol demonstrate high accuracy on benchmarks like MultiLangBench and in behavior-preserving code migration, OpenAI emphasizes that successful deployment on production systems still requires expert review and rigorous testing against business rules.

Source: Engadget

TikTok Executives Ignored Internal Child Safety Warnings

Unsealed court records reveal that TikTok executives were repeatedly warned by internal safety staff about significant child safety risks on the platform, including algorithmic rabbit holes linked to suicidal ideation and minors participating in transactional sexual content on TikTok Live. Depositions from former safety leaders like Eric Han and Reagen Maher indicate senior leadership failed to adequately address or disclose these dangers to parents, adding weight to ongoing federal multi-district litigation against major social media companies.

Source: The Verge

DeepSeek Open-Sources AI Inference Framework to Cut Costs

DeepSeek has open-sourced DeepInfer, its in-house inference framework powering the DeepSeek-V4.1-Flash model, aiming to lower compute costs, reduce energy use, and make AI model execution more affordable. By sharing production-level techniques such as distributed inference, DeepSeek allows cloud providers and enterprises to deploy its models more efficiently on their own infrastructure or adapt these optimizations for other inferencing systems.

Source: Bloomberg

Yelp Introduces AI Agent to Call Restaurants for Users

Yelp is introducing “Check for Me,” a new feature that uses a voice AI agent to call restaurants and inquire about wait times on behalf of app users. Alongside this feature, Yelp is updating its AI assistant with OpenTable reservation integration, personalized preference memory, custom reminders, and shareable lists, while also launching “Smart Lead Reply” to help small business owners automatically handle customer inquiries.

Source: TechCrunch

Spotify Launches Paid Podcast Memberships in 34 Markets

Spotify has launched paid memberships globally across 34 markets, allowing podcast creators to offer exclusive perks to their fans, such as bonus content, early access, and reduced ads, while retaining 90% of the revenue. Available to eligible creators hosting on Spotify for Creators or Megaphone, memberships are purchased via the web and expand Spotify’s monetization suite alongside advertising and partner programs to better compete with platforms like YouTube.

Source: TechCrunch