Shopify opens AI store builder as DIVD blames Zammad zero-days for its breach

Shopify introduced Canvas on 1 October, a site-building tool that lets merchants assemble a storefront by chatting with the company's AI agent Sidekick, TechCrunch reported the same day. The product renders the store's actual code in real time rather than a static preview, so merchants can test interactivity, animation and different screen sizes as the site changes.

Sidekick is not new. It has already been used to write code, build apps and customize themes, according to TechCrunch.

What changed is that Shopify gave Sidekick direct access to the theme's files and simplified the theme architecture so the agent can understand the store's structure, logic and design. The launch is narrow: third-party theme support, app blocks and extensions, markets, translations, rollouts and theme updates are all missing from the initial release, TechCrunch reported, and Canvas is desktop-only. Shopify told the outlet it plans to add these over time. The move puts Shopify alongside Wix, Squarespace, Webflow and Framer, which already sell AI-assisted site building, and the vibe-coding platforms Lovable and Replit. The difference Shopify is selling is an existing, large merchant base and an agent that edits production files rather than generating a throwaway mockup.

"We started the year at 80 sites, hit the 100th site right on our second-year anniversary as a company, and now we're nine months into the year and we've over doubled in size," Ionna CEO Seth Cutler told InsideEVs.

Security news this week moved faster than product news. The Dutch Institute for Vulnerability Disclosure said the breach of its own network was carried out by exploiting two zero-day flaws in Zammad, an open-source helpdesk and ticketing platform, BleepingComputer reported on 30 September. The flaws are tracked as CVE-2026-102489 and CVE-2026-102490 and allowed session hijacking, remote code execution and escalation to root.

DIVD had earlier described the attack as loud and messy, driven by an AI agent that chose its own next steps. Because the agent left explanations of its decisions behind, investigators could reconstruct the incident. The organization recommends Zammad users move to version 7 or take the instance offline.

Trust, and where it fails

Enterprise buyers are already nervous. Arelion's report, published on 28 September, found that 58% of senior business leaders delayed, scaled back or added safeguards to strategic initiatives in the past two years over concerns about their network provider. AI and data-driven projects were hit hardest, cited by 49% of those who experienced disruption. Security and compliance programs followed at 44%, digital transformation at 42% and cloud migrations at 38%.

The same report found a gap between stated and tested confidence: 93% of respondents said they largely or completely trust their current provider, but only 15% said they have complete confidence in that provider's ability to detect and resolve a serious network issue quickly. Asked what builds trust, leaders ranked protection against security threats first at 48%, reliability and uptime second at 44% and performance and latency third at 37%.

Vendors are selling into that gap. GTT Communications launched GTT Defense Halo on 29 September, an AI-native network defense platform it says is deployed as a dedicated instance per customer and built on its own AI factory, hosted in the United States, United Kingdom and European Union. Light Reading reported the launch. GTT says the platform identifies vulnerabilities and policy gaps, generates remediation plans and builds a stateful model of a customer's host-to-host connections.

The pitch lands against a backdrop of AI-enabled attacks. GTT's announcement points to an open letter published in late August and signed by more than 100 technology, cybersecurity and financial services companies warning that AI-enabled cyberattacks will become more widespread and that defenders have a narrow window to prepare.

Microsoft's research points the same way. The company said a malware family called NeedyMantis has been used to keep long-term access to networks that were already breached, in a small number of targeted intrusions at telecommunications organizations, universities, medical nonprofits, intergovernmental organizations and government contractors, according to The Hacker News. Its use goes back to at least October 2025. Microsoft tracked activity tied to the DAEMON Tools supply chain attack as Storm-3069 and said more than one group may be using the malware.

Agentic attacks meet agentic defenses

Training is chasing the same problem. InfoQ is running two five-week online cohorts in October 2026: AI Security and Privacy Engineering starting 26 October, and AI-Assisted Engineering starting 19 October. Katharine Jarmul, who facilitates the security cohort, said an AI security review has to follow the data and decisions across the whole system. Zichuan Xiong of Thoughtworks, who co-facilitates the engineering cohort, said the harder question for a coding agent is what it was allowed to do and how the change is known to be sound.

On the research side, arXiv posted SimTrace on 29 September, a framework that generates synthetic multimodal clickstreams from real user trajectories and a simulated twin of a web environment. The authors report it outperforms competing baselines on 7 of 8 fidelity metrics, and that augmenting real data with synthetic data improved next-action prediction accuracy by 11.0% over training on real data alone. The work is aimed at virtual clients for A/B testing, recommender development and interface evaluation, the same infrastructure that would be needed to test how an AI agent behaves on a live storefront.

Put together, the week's news reads less like a set of separate product launches and more like one argument arriving from several directions: agents are being handed more control over production systems, and the security and measurement tooling around them is still being written.