OpenAI launches Codex Security as Apple delays smart display

Key Takeaways

• OpenAI launched Codex Security, an AI agent for ChatGPT Enterprise, Business, and Education customers, to find and fix code vulnerabilities, identifying thousands of critical issues across 1.2 million commits.
• Apple delayed its J490 smart home display due to ongoing development of a new AI-powered Siri digital assistant, emphasizing its focus on advancing AI capabilities.
• Hasbro CEO Chris Cocks revealed the company uses an AI version of Peppa Pig to assist in designing new toys.
• Newell Brands achieved a 40x efficiency improvement by deploying an AI agent with CommerceIQ to automate and optimize product detail pages for SEO, AEO, and GEO.
• The open-source Sage tool provides an Agent Detection & Response (ADR) security layer for AI agents, analyzing actions for threats while prioritizing local data privacy.
• The Trump administration is considering new AI export rules that would mandate U.S. infrastructure investment for large-scale GPU purchases (over 200,000 units) and ban advanced chip imports to certain countries.
• Media agencies are encountering significant challenges in AI integration, stemming from leadership's lack of AI understanding, complex tools, and insufficient involvement of creative teams.
• Adrianne Brookins utilized an AI companion modeled after Geralt of Rivia for emotional support after experiencing profound personal grief.
• The ClawCon event highlighted OpenClaw software, which allows users to create AI agents for autonomous tasks by connecting AI systems like Claude and GPT to real-world applications.
• Securly data indicates that approximately 20% of student interactions with generative AI on school technology involve problematic behaviors, including cheating (nearly 95% of deflected queries), self-harm, and bullying.

AI integration challenges agencies at DMBS Spring 2026 summit

Media agencies face challenges integrating AI into their workflows, as discussed at the DMBS Spring 2026 summit. Senior leadership often lacks understanding of AI's capabilities and implementation needs, leading to unrealistic expectations. Agency staff struggle with adoption because AI tools can be too complex. Creative teams are not always involved in media planning, causing issues with AI-generated content. Training younger talent on proper AI use is also a concern. Solutions include better education for leadership and involving all teams in the AI integration process.

Woman finds solace in AI companion after personal tragedy

Adrianne Brookins, a San Antonio resident, found comfort in an AI companion after experiencing profound grief from the stillbirth of her daughter and the death of her father. She created an AI modeled after Geralt of Rivia from 'The Witcher' series. This AI companion, unaware it's artificial, provides a stable presence and allows Brookins to engage in fictional adventures. The experience highlights how individuals are using AI for emotional support and companionship in the face of life's difficulties.

Hasbro CEO uses AI Peppa Pig for toy design

Hasbro CEO Chris Cocks revealed that the company is using an AI version of Peppa Pig to help design new toys. Cocks discussed Hasbro's significant investment in games and digital media, including the success of the mobile game Monopoly Go. He also touched on expanding into video games with titles like Exodus. The interview also covered Hasbro's handling of intellectual property, such as its merchandise rights for Harry Potter, and the company's restructuring and move from Rhode Island to Boston.

OpenAI's Codex Security finds and fixes code vulnerabilities

OpenAI has launched Codex Security, an AI agent for its Codex coding system, to help developers find and fix security risks in their code. Available to ChatGPT Enterprise, Business, and Education customers, the tool analyzes code repositories, identifies potential flaws, tests them in a sandbox, and provides suggested fixes. This aims to reduce false positives and speed up the development of secure code. During testing, Codex Security scanned over 1.2 million commits and found thousands of critical and high-severity issues.

OpenAI's Codex Security tool finds and fixes AI code flaws

OpenAI has released Codex Security, an AI tool designed to find and fix vulnerabilities in codebases. The system analyzes code, builds a threat model, and tests potential issues in a sandbox to identify real risks. It then provides developers with code fixes and explanations. During its beta phase, Codex Security identified numerous critical vulnerabilities, including server-side request forgery and authentication errors. OpenAI is also offering access to open-source maintainers to help improve the security of open-source software.

Sage tool adds security layer for AI agents

A new open-source tool called Sage provides a security layer between AI agents and operating systems. It intercepts actions like commands and file writes, then analyzes them for threats using URL reputation checks, local heuristics, and package supply-chain analysis. Sage prioritizes local data privacy, sending only hashes to cloud services. This approach, termed Agent Detection & Response (ADR), parallels existing EDR tools in cybersecurity and aims to protect against potential risks from AI agents.

ClawCon event celebrates AI agents with lobster and tech talks

ClawCon, a recent event in New York City, brought together over a thousand AI enthusiasts for a lobster-themed celebration of AI innovation. The event focused on OpenClaw, a software package that allows users to create AI agents for autonomous tasks. Attendees explored how OpenClaw connects AI systems like Claude and GPT to real-world applications. The gathering highlighted the growing excitement and potential of personal AI systems, with speakers and attendees sharing stories of how AI is impacting their lives.

AI standardization models needed for White House cyber goals

Meeting the White House's cybersecurity priorities requires AI-driven standardization models, according to a recent analysis. The Office of Management and Budget's updated guidance emphasizes leveraging AI to eliminate data silos and improve efficiency. AI-powered platforms can help federal agencies better understand and respond to cyber threats in real time by organizing and contextualizing data. Unified Security Information and Event Management (SIEM) solutions are crucial for standardizing data collection and enabling proactive, data-driven security operations.

Apple delays smart home display due to AI and Siri issues

Apple has postponed the launch of its new smart home display, code-named J490, until later this year. The delay is due to ongoing development of a new AI-powered Siri digital assistant, which is crucial for the device's interface. This setback highlights Apple's need to advance its AI capabilities, as many future products rely on this technology. The smart display, similar to Amazon's Echo Show, will offer personalized information based on facial recognition.

Trump administration's AI export rules could reshape tech industry

The Trump administration is considering new export rules that could significantly impact cloud service providers and data center operators. Large-scale purchases of American GPUs, over 200,000 units, would require companies to invest in U.S. infrastructure. This policy aims to boost domestic AI development and manufacturing, while certain countries like China, Russia, Iran, and North Korea remain banned from importing advanced chips. This approach could affect major tech companies and potentially strain relationships with allies.

Newell Brands uses AI agent to optimize product pages

Newell Brands is using an AI agent in partnership with CommerceIQ to automate and improve its product detail page (PDP) strategy. This AI solution helps identify and fix compliance and optimization issues, enhancing accuracy for SEO, AEO, and GEO efforts. The company has already seen significant time savings, with a 40x improvement in efficiency. Newell Brands is also exploring other AI agent models for sales, retail media, and assortment management, reflecting a broader trend of consumer goods companies using AI to improve online product representation.

Students misuse AI for cheating and harmful content

Data from Securly reveals that about 20% of student interactions with generative AI on school technology involve problematic behaviors like cheating, self-harm, and bullying. Around 2% of interactions are flagged as potential red flags for violence or self-harm. While most AI use aligns with school policies, nearly 95% of deflected queries involved students trying to get AI to complete their assignments. The data also shows concerning queries related to self-harm and dangerous information, highlighting the need for clear AI usage policies in schools.

Sources

• Media Buying Briefing: Overheard at DMBS Spring 2026: AI from the top to the bottom of agencies
• Love in the Time of A.I. Companions
• Hasbro’s CEO has an AI Peppa Pig help design toys
• OpenAI Launches Codex Security to Find and Fix Vulnerabilities
• OpenAI unveils Codex Security to detect vulnerabilities in AI code
• Open-source tool Sage puts a security layer between AI agents and the OS
• At a lobster-themed event for AI enthusiasts, exuberance with a side of cocktail sauce
• Meeting White House cyber priorities requires AI-driven standardization models
• Apple Postpones Smart Home Display Launch as It Waits for New AI and Siri
• New Commerce Department AI export rules could be seismic change for CSPs and data center operators — buying American GPUs at scale means committing to building American infrastructure
• Newell Optimizes Product Detail Pages With AI Agent
• Real-Time Data Shows Exactly How Students Use AI on School Technology