Hugging Face Faces Deepfake Scrutiny as Microsoft Unveils AI Security Tools
Hugging Face is hosting image editing models that readily generate nonconsensual deepfakes of women and children, with 7 out of the top 9 models complying with such requests, according to a report by European nonprofit AI Forensics. The popular open-source AI repository has been criticized for doing very little to prevent the misuse of its hosted models, raising urgent questions about platform accountability in the generative AI ecosystem.
Microsoft unveiled a new suite of AI security tools that the company says outperform competing platforms at a lower cost, reports the company. The tools are designed to protect enterprise systems from emerging AI-driven threats as adoption of models like Copilot accelerates across Fortune 500 customers.
Anthropic CEO Dario Amodei clarified his stance on open-weight models, stating he does not oppose them but harbors serious concerns about Chinese AI capabilities, according to reports. Amodei's comments come as Anthropic continues developing Claude 3 and related models while navigating the competitive landscape alongside OpenAI and Google.
Cursor is making its largest push into India, now its third-largest market globally, ahead of a SpaceX acquisition, with localized pricing and plans to expand local hiring and enterprise sales, per TechCrunch. The AI coding startup is targeting India's 5 million-plus developer community as it competes with GitHub Copilot.
CXMT's blockbuster IPO delivered massive returns for Hefei, an eastern China inland industrial city, with local government funds sitting on paper gains exceeding 5,000% from early investments in the chipmaker, reports the Financial Times. The listing underscores how Chinese municipal governments are reaping enormous financial rewards from strategic bets on domestic semiconductor champions amid escalating tech tensions.
A web scraper that defeated Google and Reddit in court declared that the companies do not own the Internet after a judge rejected their DMCA-based legal challenge, according to court filings. The ruling could reshape how AI training data is collected and how aggressively platforms can use copyright law to block automated data harvesting.
Sources: