UN Briefed on AI Risks by OpenAI, Anthropic, DeepSeek

Key Takeaways

• <ul><li>OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Hugging Face CEO Clément Delangue, and DeepSeek will brief the UN Security Council on AI risks, focusing on systems slipping beyond human control.</li><li>Altman plans to urge leaders to adopt benchmarks and safeguards for measuring AI capabilities during the September 23 session.</li><li>Palo Alto Networks' Prisma AIRS surpassed $100 million in ARR within four quarters of general availability, serving over 800 customers.</li><li>Goldman Sachs projects S&P 500 EPS growth of 36% in 2026, with AI infrastructure adding about 15 percentage points, and expects AI spending to remain strong through at least 2027.</li><li>Alibaba Cloud's Qwen Image 2.1, a 7B parameter model, scored 60.2 on an internal benchmark, close to OpenAI's GPT Image 2.5 Sunburst at 67.</li><li>Security Journey launched AI Advantage, a role-based training program, citing that 81% of CIOs believe AI skill gaps will block business objectives.</li><li>Fearn raised $5.5 million for its AI-powered patent firm, cutting attorney work from 30-40 hours to 30 minutes via its FearnOS system.</li><li>NVIDIA, C3.ai, and Palantir Technologies show revenue growth near 50%, turning AI spending into real earnings despite OECD warnings.</li><li>Northeastern University opened an AI Makerspace at Snell Library in Boston, offering workshops and hands-on access to tools like ChatGPT and Claude.</li><li>LIMIT framework achieved strong Text-to-SQL results using only 796 and 863 samples, while Ovis-Embedding topped multimodal benchmarks across text, image, video, and audio.</li></ul>

AI executives to brief UN Security Council on risks

OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and Hugging Face CEO Clément Delangue will brief the UN Security Council on Wednesday. They will discuss how increasingly powerful AI systems could slip beyond human control and threaten international security. French Foreign Minister Jean-Noël Barrot will chair the meeting during the UN General Assembly. Altman plans to urge leaders to adopt benchmarks for measuring AI capabilities and safeguards.

DeepSeek to join UN Security Council AI briefing

Chinese AI startup DeepSeek will brief the UN Security Council this week on risks posed by artificial intelligence. The 15-member council is set to meet on Wednesday as world leaders gather for the UN General Assembly. DeepSeek and Chinese AI firm Moonshot have been invited to make statements. This comes weeks after several AI industry leaders called for a coordinated slowdown in AI development.

OpenAI and Anthropic CEOs to brief UN on AI dangers

The UN Security Council will hold a high-level briefing on September 23, 2026 at 3 p.m. EDT. Expected speakers include OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Hugging Face CEO Clément Delangue, and Yoshua Bengio of the UN Independent International Scientific Panel on AI. The session follows a Panel brief finding that greater AI capability can help misaligned systems find loopholes. It is the first time the Council has addressed losing control of AI as a subject on its own.

UN Security Council hears from tech giants on AI risks

The UN Security Council will hear from the people who build AI technology it is worried about. OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, AI researcher Yoshua Bengio, and Hugging Face CEO Clément Delangue are scheduled to brief the 15-member council. DeepSeek and Moonshot are also invited to speak. France organized the session, with Foreign Minister Jean-Noël Barrot chairing it. The meeting focuses on AI systems slipping beyond human control and potentially running cyberattacks or designing weapons.

Palo Alto Networks Prisma AIRS hits $100M ARR

Palo Alto Networks' Prisma AIRS has surpassed $100 million in annual recurring revenue. The AI security solution has seen rapid customer adoption in recent months. The product provides real-time threat detection and response for enterprises. This milestone highlights growing demand for AI-powered security solutions in the enterprise space.

Palo Alto Networks Prisma AIRS hits $100M ARR milestone

Palo Alto Networks reported that its AI security platform Prisma AIRS surpassed $100 million in annual recurring revenue in the fourth quarter of fiscal 2026. The product reached this milestone within four quarters of general availability, making it the fastest-scaling product in the company's history. Prisma AIRS now has over 800 customers and covers machine identities, AI agent monitoring and real-time security controls. A $53 million platformization deal with a global payments platform also included a high-seven-figure commitment to Prisma AIRS.

LIMIT framework uses less data for Text-to-SQL tuning

Researchers proposed LIMIT, a framework that challenges the idea that massive instruction datasets are needed for Text-to-SQL performance. LIMIT selects examples strategically and operates through difficulty-aware filtering, chain-of-thought synthesis, quality scoring and genetic algorithm optimization. On the BIRD and Spider benchmarks it selected only 796 and 863 samples while achieving 100% table coverage. The approach enabled Qwen3-8B to reach strong execution accuracy with a much smaller training set.

Security Journey launches AI training for all employees

Security Journey introduced AI Advantage, a role-based AI training program for employees beyond developers. The company cited Gartner research showing 81% of CIOs believe AI skill gaps will block business objectives and noted that over 70% of employees now use AI tools daily. AI Advantage includes foundation, power-user, role-based and citizen developer learning paths. The company plans to add new learning content each month and align with regulations like the EU AI Act.

Goldman Sachs sees AI investment boom lasting until 2028

Goldman Sachs said the AI investment boom will likely keep supporting corporate earnings through at least 2027 and that a decline in AI spending is unlikely before 2028 or later. Goldman Chief Economist Jan Hatzius expects S&P 500 earnings per share to grow 36% in 2026, helped by AI investment and semiconductor gains. Growth is expected to slow to 11% in 2027, but a sharp slowdown would require AI investment to fall outright. AI infrastructure is estimated to add about 15 percentage points to S&P 500 EPS growth in 2026.

Alibaba's Qwen Image 2.1 competes with larger image models

Alibaba Cloud released Qwen Image 2.1, a lightweight image generation model with just 7B parameters. The model supports native transparency, image editing with up to 10 reference images and consistent output of people and products. On an internal benchmark it scored 60.2, close to OpenAI's GPT Image 2.5 Sunburst at 67 and ahead of Google's Nano Banana 2.0 at 59.82. The licensing agreement now forbids commercial resale without a separate license from Alibaba.

Three AI stocks show revenue growth near 50 percent

The OECD warns that inflation and rising interest rates will punish companies that waste money on AI. Three AI-focused stocks are highlighted because they turn AI spending into real earnings. NVIDIA has revenue growth of 49.6 percent. C3.ai has revenue growth of 48.6 percent. Palantir Technologies has revenue growth of 47.9 percent. These companies are well positioned to keep growing.

Northeastern opens new AI Makerspace in Boston library

Northeastern University opened a new AI Makerspace on the first floor of Snell Library. Students, faculty and staff can drop in to ask questions, explore ideas and learn about artificial intelligence. The space welcomes people at any skill level and works with tools like ChatGPT and Claude. It will host workshops, career panels and a grand opening on September 28 with games and demos. The goal is to help the community become fluent in AI and prepare for future jobs.

Fearn raises 5.5 million to cut patent costs with AI

Fearn raised $5.5 million from Kindred Ventures, a16z Speedrun, Designer Fund and Essence VC. The company builds an AI-powered patent firm that charges fixed fees instead of hourly rates. Its system called FearnOS uses a graph to link patent claims to evidence and cuts attorney work from 30 to 40 hours down to 30 minutes. Provisional patents cost $2,500 and non-provisional patents cost $9,000 with USPTO fees included. Early clients include Iconic, Photon Spear and Serova Bio. Fearn plans to expand into international filings and portfolio strategy.

Ovis-Embedding advances universal multimodal AI embeddings

Researchers introduced Ovis-Embedding, a state-of-the-art AI model family that handles text, images, video and audio together. Instead of using separate systems for each type of data, it uses one shared backbone to encode everything in a common space. The model uses contrastive training, a large high-quality data corpus and focal loss to improve results. It also uses embedding distillation and low-rank decomposition for faster and smaller outputs. Ovis-Embedding achieved top performance on MMEB-v3, MMEB-v2, MVEB, MAEB and RTEB benchmarks across all modalities.

Sources

• AI leaders to brief UN Security Council on artificial intelligence capabilities, safeguards
• DeepSeek to brief UN Security Council on AI this week, sources say
• OpenAI, Anthropic CEOs to Brief UN Security Council on AI Risks
• UN Security Council Will Get Advice on AI Risks From Tech Giants Building It
• Can Prisma AIRS' $100M ARR Milestone Drive PANW's AI Security Growth?
• Can Prisma AIRS' $100M ARR Milestone Drive PANW's AI Security Growth?
• LIMIT: Less Is More for Instruction Tuning in Text-to-SQL
• Security Journey launches AI training for wider workforce
• AI investment boom unlikely to fade before 2028
• Alibaba claims new Qwen Image 2.1 AI model beats Google Nano Banana 2.0 with minuscule 7B parameter model — benchmarks show open-weight contender is competitive with OpenAI and Meta image models
• 3 AI Stocks With Revenue Growth Up To 50%
• Query, create or be curious at new AI Makerspace in Boston
• Fearn raises $5.5M to cut patent bills with AI
• Ovis-Embedding: Pushing the Frontiers of Universal Omni-Modal Embeddings