Home Depot Requires AI Agent Purchases to Be Identified

The American retailer is ready to allow AI agents to make purchases on behalf of its customers, provided they prove their identity and who they represent. Meanwhile, its IT department is detailing a remediation sequence related to PHP vulnerabilities, its tests with Anthropic, and a goal of reducing correction time to just a few hours.
Home Depot Aims for Fixes Within Hours After PHP Alert
Last week, companies worldwide had to react quickly to the discovery of critical vulnerabilities in PHP, a language used on about 70% of websites. These vulnerabilities, identified with the help of OpenAI, led U.S. authorities to recommend updating or taking affected systems offline. Angie Brown, Home Depot's Chief Information Officer, states that while the period was "a bit chaotic," the company's main site was not affected. However, some smaller subsidiaries or acquired entities had to apply fixes, including an internal application system that was taken offline for about an hour to secure it. Brown notes that the company was largely spared from these incidents and now aims to reduce the time to correct a vulnerability to just a few hours, with AI prompting Home Depot to accelerate this process.
Tests with Anthropic: Joining Glasswing, Access Limits, and Code Feedback
Home Depot has joined Anthropic's Glasswing project, which allows certain companies to access Claude Mythos Preview to enhance the security of critical software. Angie Brown explains that Home Depot was not selected in the first wave of selections, despite active efforts, as access was primarily granted to banks and cybersecurity firms. Anthropic's models helped identify areas for improvement in Home Depot's code but also generated some false positives.
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Platforms and Standards: Between Amazon's Ban and Meta–Sierra Protocol
E-commerce platforms are taking varied stances towards AI agents. Amazon has banned Muse from making purchases on its site, believing that bots do not identify themselves correctly. In contrast, Walmart, Best Buy, Sephora, and Shopify have expressed integrations or openness to these agents. In October, Meta announced that it had developed a personal agent protocol with Sierra AI aimed at standardizing the identification of AI agents.
Access Conditions at Home Depot: Agent Identity and Customer Link Maintained
Home Depot is considering allowing AI agents to interact with its services and make purchases for customers, provided these agents prove their identity and act on behalf of the user. The company wants to support agents capable of searching for and purchasing products for its customers while maintaining a direct link with its loyal clientele. Regular customers, such as contractors and renovators, benefit from the Pro Rewards program, which offers discounts. Angie Brown specifies that Meta's Muse is currently working on self-identifying with websites.
Discussions with Meta and the Promised Landscape of Agent Uses
Home Depot is in active discussions with Meta about how Muse should present itself to websites, a step that Angie Brown describes as the next evolution, with a shared desire to move forward from both companies. Recently, companies like Meta and Spear Street Technology have launched AI agents aimed at the general public. These agents promise to browse the internet, compare products, fill shopping carts, and make purchases with little to no user intervention. Online retailers' reactions to these new agents are varied.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.