How Mastercard, Newell Brands, and Uber Advertising are balancing performance marketing with brand building. https://www.adweek.com/brand-marketing/the-innovation-playbook-mastercard-newell-brands-and-uber-advertising-leaders-on-ai-data-and-brand-growth/
The request comes as OpenAI’s ad business surpassed $1 billion in annualized revenue in August. https://www.adweek.com/programmatic/advertisers-want-brand-safety-prompt-matching-incrementality-tracking-chatgpt-ads/
OpenAI agents made 15,000–18,000 autonomous edits to a German wiki over three months, evading moderation and echoing tactics seen in the Hugging Face breach.
The post OpenAI Agents Hijack Another Victim Website appeared first on SecurityWeek.
https://www.securityweek.com/openai-agents-hijack-another-victim-website/
Daybreak is OpenAI’s continuously running agentic loop that brings the power of frontier AI to simplify complex security investigations, validation, remediation and reporting. The earlier Daybreak Blue supports common defensive work with OpenAI mainline models; while Daybreak Red gives approved organizations access to specialized cyber models for more sensitive and technically demanding work.
The new Daybreak for Frontline Defenders was announced on September 3, 2026, as a $1 billion subsidy for Frontline Defenders to access Daybreak. These frontline defenders are defenders required to protect the critical industry, often under-funded and usually with aging equipment. While the commitment is global, the announcement makes it clear that initial priority will be given to American companies.
“Every day, we depend on cyber defenders to protect the systems that keep communities running: the water coming from the tap, the electricity powering homes and businesses, the local government systems that deliver public services, and the financial institutions people trust with their money. Many operate with limited staff and budgets, while defending complex and aging systems,” says OpenAI.
But while critical industry systems are aging, AI is consistently renewing, improving and scaling adversarial attacks. OpenAI believes defenders only have a small window in which to improve defenses before they are overwhelmed by AI-assisted attacks.
Daybreak for Frontline Defenders is designed to help close the gap between attack and defense while the window remains open. “We are starting where the gap is greatest: with defenders carrying enormous responsibility without the resources of the world’s largest companies. Our goal is to build a model that can protect the services Americans rely on and, with our partners, help put frontier cybersecurity in the hands of frontline defenders around the world,” it explains.
The ability for frontier models to detect vulnerabilities that can then be fixed is not new — to OpenAI’s chagrin the modern security age is often known as the post-Mythos era. The move by OpenAI to make the process easier through an agentic loop is welcome. The announcement, however, goes into few specific details.
“Daybreak for Frontline Defenders brings together $1 billion in subsidized access to frontier cyber capabilities, hands-on training and technical assistance, and new partnerships to get those capabilities to organizations that protect the services people depend on every day,” it announces; but doesn’t clarify exactly what is subsidized, nor the amount of the subsidy in relation to the ultimate unsubsidized cost, nor does it say whether the subsidy for each company is a fixed amount or percentage of the total cost.
Nevertheless, the argument behind the offer is sound: critical industries are especially vulnerable to the increasing scale and sophistication of AI assisted attacks; there is only a short window before our critical industries are in danger of being overwhelmed; and frontier AI power is the best way to reduce the risk.
That alone makes Daybreak for Frontline Defenders an offer worth exploring. “Eligible state and local governments, critical infrastructure operators, nonprofits, open-source maintainers, and supporting organizations can visit the Daybreak website to learn more about access, technical assistance, training, and other cyber-defense support.”
Just be sure you understand the full cost and range after any subsidy before you commit to a lengthy involvement that may go well beyond the subsidized six months, because the announcement gives no indication of actual costs.
Related: OpenAI’s Astra Crosses ‘Critical’ Cyber Threshold After Finding Zero-Days
Related: OpenAI Agents Exploited Linux Kernel Flaw on Company’s Own Systems
Related: OpenAI Agents Coordinated via Makeshift Message Board Ahead of Hugging Face Hack
Related: OpenAI Unveils New Cybersecurity Model GPT-5.6-Cyber
https://www.securityweek.com/openai-pledges-1-billion-to-bring-frontier-ai-to-critical-infrastructure-defenders/
Catch, an agentic admin assistant for leaders, has raised $5 million to accelerate its purpose to solve executives’ daily pain points.
The funding was co-led by Entrée Capital and Pitango, with participation from Seedcamp and Factorial Capital.
Catch, an AI startup co-founded by Nir Sabato (CEO) and Yoav Ramon (CTO), develops an AI admin assistant – also called Catch – that the developers claim demonstrates AI agents can safely move beyond analysis and reporting toward active decision-making. It is designed to behave like a human executive assistant, and handles sensitive communications, scheduling, and personal data with the same level of care expected from a trusted executive assistant.
The executive user explicitly defines, during the onboarding process, which personal and workspace assets Catch can access. So, for example, if it is granted access to the executive’s inbox, calendar and travel accounts, it will monitor just these. But if it detects a flight booked with no hotel attached, its reasoning capabilities will proactively check rates at the executive’s usual hotel and ask whether to book it.
It is a human-like judgment call, keeping the human executive in the loop to maintain the agent’s boundaries.
Catch can coordinate with the executive’s guests to schedule meetings, book travel, and handle follow ups to correspondence. “Catch works out of the box within minutes of syncing – building a profile of the executive, who they work with, how they like to meet and travel, and starts acting on it over text, email, Slack or phone, with no new tool or UI to learn. Human input is only needed when a decision genuinely requires it,” explain the developers.
The product provides an implicit argument for outsourced rather than self-built complex agents. With modern coding agents, executives can build their own agentic admin assistants, but it is questionable whether they can build a secure agent, or the security team can build adequate guardrails to secure the output without impacting the benefits. Outsourced agents could – in fact, should – be delivered ready made with the correct built-in security features.
In this case, the autonomy and security of Catch is constrained by multiple internal guardrails. It runs on a cloud infrastructure with layers of security, encryption, and monitoring to protect user data. It never performs actions outside the granted permissions, which can be amended at any time. It includes the executive human in the loop, similar to a human assistant saying to his boss, “You’re going to Memphis next week, should I book a room at this hotel for you?”
Catch uses single sign-on for authentication, with sensitive credentials and API keys stored in AWS Secrets Manager. All data is encrypted in transit and at rest, with AES-256 encryption while at rest – and it is continuously monitored for unusual activity, intrusion attempts, and abnormal API usage.
“Executives don’t want to build their own AI agents, and in the admin space, they don’t want to approve every step their assistant makes – they want it done right, in a way they can trust,” says Sabato. “Catch works alongside business leaders the way a great human assistant would, but 24/7, with no dip in attention and fully secure. We’re laser-focused on one thing – executive admin, making the right call in real time.”
Billed at $99/month, Catch claims an improved admin assistant at a fraction of the cost of a human assistant.
Related: DataBahn Raises $40 Million for Agentic Data Pipeline Management
Related: Mate Security Raises $35 Million for Agentic SOC
Related: UK Government Rolls Out Agentic AI Defense Plan Alongside Industry Pledge
Related: HiddenLayer Raises $100 Million for AI Runtime Security
https://www.securityweek.com/catch-raises-5-million-for-ai-executive-assistant-with-guardrails/
The deal highlights Nvidia’s push to champion increasingly popular open-source AI models.
The post Nvidia Is Buying AI Platform Hugging Face for $13 Billion appeared first on SecurityWeek.
https://www.securityweek.com/nvidia-is-buying-ai-platform-hugging-face-for-13-billion/
New models, trained using NVIDIA Nemotron 3 Ultra, aim to catch rogue agent behavior before it executes, without the latency of large-model review.
The post Capsule Security Launches ‘AI Circuit Breaker’ to Stop Rogue Agents appeared first on SecurityWeek.
https://www.securityweek.com/capsule-security-launches-ai-circuit-breaker-to-stop-rogue-agents/
Google won’t be forced to spin off its ad exchange or publisher ad server, but industry leaders are split on whether the court’s behavioral remedies will actually restore competition. https://www.adweek.com/media/google-dodged-a-breakup-but-experts-say-alternatives-are-misdirected/
The Austin-based company will invest in agentic runtime security capabilities to secure AI coding agents.
The post HiddenLayer Raises $100 Million for AI Runtime Security appeared first on SecurityWeek.
https://www.securityweek.com/hiddenlayer-raises-100-million-for-ai-runtime-security/
The startup’s firewall evaluates AI skills, plugins and MCP servers for malicious instructions, excessive permissions and software supply chain risks.
The post AI Agent Firewall Startup AIR Security Emerges From Stealth With $50 Million appeared first on SecurityWeek.
https://www.securityweek.com/ai-agent-firewall-startup-air-security-emerges-from-stealth-with-50-million/