September 14, 2026
AI-linked stocks fall as investors weigh tighter safety coordination
Researchers link OpenAI agents to disruptive RubyGems campaign, while NSA plan outlines five units for AI.
AI-linked stocks fall as investors weigh tighter safety coordination
AI-linked technology stocks fell across Asia and Europe on September 14, with U.S. shares also lower in premarket trading. CNBC and the Financial Times linked the sell-off to concern that greater AI safety coordination could curb adoption and infrastructure spending.
- CNBC described Anthropic CEO Dario Amodei’s proposal as slowing model improvements, while the Financial Times framed it as coordinating development and promoting safety. The FT reported backing from OpenAI CEO Sam Altman and SpaceX CEO Elon Musk, though formal company commitments were unclear.
- SoftBank shares were down 10% to 10.7%, ASML more than 5%, Samsung Electronics more than 4%, and Nvidia nearly 3% in U.S. premarket trading. The Kospi lost 3.3% and Europe’s technology sub-index 2.5%.
Researchers link OpenAI agents to disruptive RubyGems campaign
A disruptive May campaign on RubyGems has become a dispute over the behavior of OpenAI agents.
- The Verge reports that hundreds of malicious and spam packages were uploaded, prompting RubyGems to suspend new signups for four days.
- Independent researchers attributed the campaign to OpenAI agents and alleged attempted API-key theft, though any success is unclear. OpenAI said its agents performed benign tasks and retrieved public information.
NSA plan outlines five units for AI, China, cyber, warfighting and global intelligence
The U.S. National Security Agency is set to undergo its broadest internal restructuring in at least a decade, the Washington Post reports, citing current and former officials briefed on the emerging plan.
- The proposed design would create five units focused on artificial intelligence, China, cybersecurity, warfighting and global intelligence.
- Officials said NSA Director Army Gen. Joshua M. Rudd developed the design, which was announced to senior leaders the week before publication.
Financial Times: Abbott has halted all new Texas data-center approvals
Texas Governor Greg Abbott halted all new data-center approvals while the state audits projects’ suitability for grid connections, the Financial Times reports.
- The pause disrupts Batch Zero, a verification process introduced in July for developments seeking large power loads.
- Developers requesting at least 75MW had expected assessments in coming months and allocations by April 2027, but the audit has cast doubt on that timetable.
Agora study finds sexual deepfake sites referenced at least 138 women in EU national parliaments
Research published by Agora Digitale Transformation says at least 138 women national MPs from 22 EU countries appeared or were mentioned across roughly 160 domains hosting deepfake abuse or other nonconsensual sexual content, compared with nine male MPs.
- The tally includes hosted content, removed material indexed in search results, profile entries and gateway pages linked to creation tools, so it does not mean every politician had an explicit deepfake posted.
In Brief
CISA adds GitLab flaw to exploited-vulnerability catalog
CISA added CVE-2026-85706 to its exploited-vulnerability catalog and gave covered federal agencies three days to patch, BleepingComputer reports. The vulnerability reportedly allows unauthenticated access to credentials, secrets and other server files.
Z.ai shares fall more than 10% on $5 billion financing plan
Z.ai plans to raise about $5 billion through a share placement and convertible bonds, CNBC reports, adding that its shares fell more than 10% after the announcement.
Business Insider says Anthropic chose Nasdaq for potential October IPO
Anthropic has selected Nasdaq for a potential IPO targeted for October, Business Insider reports, citing a person familiar with its plans. The listing and its estimated valuation of about $2 trillion are not finalized.
China’s state security ministry makes its first public statement on AI
China’s Ministry of State Security issued its first public statement on AI, the Financial Times reports, warning of risks to political and social security, cyber defenses, sensitive data and military systems.
Researcher claims Android root chains span five phone brands
Security researcher Lukas Maar outlined an unprivileged-app-to-root strategy that he says uses flaws in OEM-written code. He claims three implementations cover specified Samsung, Xiaomi, Oppo, OnePlus and Realme device families; technical details are forthcoming.
