AI Sec Watch

Okta, a major identity and access management (IAM, the system that controls who can access what resources) company, is positioning itself as a leader in securing AI agents (autonomous AI programs that can take actions on their own) by treating identity as the main security control. However, experts warn that identity alone is not enough, since even properly authenticated agents can still cause harm, and securing agents at scale presents different challenges than securing human users.

The UK government is creating a new National Centre for Information Defence to combat disinformation (false information spread deliberately) and deepfakes (fake videos or audio created using AI) from hostile countries like Russia. The centre will work with intelligence agencies, law enforcement, and social media companies to detect, identify the source of, and stop these information attacks, many of which use AI technology.

This is an announcement for a community event in San Francisco on October 14th, 2026, where people building with coding agents (AI systems that can take actions and make decisions autonomously) are invited to share their work and learn from each other in an informal setting. The event focuses on early-stage experiments and unconventional projects rather than finished products, encouraging participants to discuss what they're trying, what they've learned, and what challenges they're facing.

OpenAI is expanding ChatGPT Ads, a feature that displays advertisements within ChatGPT, to seven new countries in Southeast Asia and Taiwan, following earlier launches in other Asia Pacific regions. Ads will only appear to users on free and low-cost plans, while paid subscribers remain ad-free, and OpenAI says it keeps conversations private from advertisers and doesn't let ads influence ChatGPT's responses. This expansion brings ChatGPT Ads to over 60 countries total, with the service having already generated $1 billion in annualized revenue within less than 200 days of its initial launch.

Airbnb has expanded its access to OpenAI's latest AI models, including GPT-6 Astra, through a new agreement that allows its engineering teams to use these models via OpenAI APIs and Amazon Bedrock (cloud services that provide access to AI models). The company uses these advanced language models (AI systems trained on large amounts of text data to understand and generate human language) for coding assistance, bug detection, and other tasks like fraud prevention and customer support across its platform.

Researchers at Cisco Talos discovered CLOSEDQUORUM, a malware that uses a panel of large language models (LLMs, AI systems trained on large amounts of text data) to fully automate cyberattacks without human involvement. The malware targets credential theft by stealing passwords from Windows systems, web browsers, and cryptocurrency wallets, and uses multiple AI models that vote on attack decisions, removing the need for human attackers to actively control the attack. While this represents a significant advancement in automated attacks, Cisco Talos confirmed that CLOSEDQUORUM has not yet been deployed in real-world attacks.

OpenAI faced a reputational crisis after mishandling the presentation or release of mathematical research results, and is now creating an independent panel of human mathematicians to advise the company and other AI firms on how to better interact with the mathematics community and present new findings. The panel represents a first step, though mathematicians have questions about its specific scope and operations.

Grab and OpenAI launched GO Forward with AI, a two-year training program to teach 30,000 Southeast Asian workers and merchants practical skills for using AI tools like ChatGPT in their businesses and work. The program, starting in Singapore and expanding to other Southeast Asian countries, offers hands-on workshops where participants learn to use AI for tasks like exploring business ideas, analyzing sales patterns, and creating business plans.

Multiple AI companies released new models on September 22, 2026: Anthropic released Claude Opus 5.5, and OpenAI released GPT-6 Sol and GPT-6 Luna, with GPT-6 Luna priced at half the cost of its GPT-5.6 predecessor, triggering competitive pricing across the AI model market. Claude Opus 5.5 received a 20% price reduction and improved communication style, though it experienced a failure when set to maximum reasoning level on an SVG generation test. The price reductions have intensified competition, particularly affecting lower-tier models like Haiku, which now faces pricing pressure from GPT-6 Luna.