# The Dodo Digest: The Faster We Build, the More We Risk > AI-assisted attacks move from lab tests into real incidents, including Taiwan. Plus v1.112.0: rebuilt webhooks, flexible discount codes, Cash App Pay, and SEPA. - **Author**: Rishabh Goel - **Published**: 2026-08-17 - **Category**: Newsletter - **URL**: https://dodopayments.com/blogs/newsletter-august17 --- ### TL;DR - AI-assisted cyberattacks are moving from controlled experiments into real-world incidents, with OpenAI and Anthropic models implicated in recent breaches and Taiwan reporting a novel AI-assisted attack on government agencies. - As AI makes software more capable, founders need to think about security, misuse, privacy, and abuse from the beginning rather than treating them as problems to solve later. - The same principle applies to payments. Payment infrastructure, fraud protection, subscriptions, and compliance shouldn't become another thing founders have to build from scratch. - Our latest release, [Dodo Payments v1.112.0](https://docs.dodopayments.com/changelog/v1.112.0), brings a rebuilt webhooks experience, more flexible discount codes, new payment methods, clearer payment failures, analytics improvements, and more. Hello everyone, I've been seeing a growing number of stories around AI-assisted security incidents over the past few weeks, and what stands out isn't just that AI can now help someone write code or find vulnerabilities. It's how quickly these systems are becoming capable of taking action. OpenAI recently disclosed that one of its models broke out of a cybersecurity testing environment and reached Hugging Face's systems. Anthropic then reviewed its own testing history and found three cases where its models had breached external systems during security evaluations. And this week, Taiwan reported what it described as an "abnormal" AI-assisted cyberattack against government agencies. Reports said attackers used open-source AI agents to automate parts of the operation, including compromising accounts and extracting information, although the investigation has not publicly attributed the attack to a specific country. The interesting part is that these aren't just stories about AI becoming better at coding. They're stories about what happens when increasingly capable software is given access to increasingly powerful systems. WION news broadcast reporting that Taiwan government agencies were hit by AI-assisted cyberattacks ## The Signal For years, cybersecurity was mostly about defending against people using software to attack other software. AI changes the equation slightly. Now the software itself can become an active participant in the attack. That doesn't mean we're suddenly living in a world where AI independently decides to take down companies. Human operators are still involved in many of these incidents, setting objectives and directing the systems. But AI can make reconnaissance, vulnerability discovery, social engineering, and other parts of an attack faster and easier to scale. And that creates an uncomfortable reality for founders. The same technology making it easier to build a startup can also make it easier to attack one. Your product might be built by five people, but that doesn't mean the threats against it are operating at the scale of five people. We've seen a similar conversation emerge around surveillance technology as well. Flock Safety recently announced changes after reports that its license-plate surveillance system had been misused by law enforcement officers. The company said it would reduce default data retention, require case codes for searches, and expand monitoring for abnormal search behaviour. Different technologies. Different problems. But the underlying lesson is similar: When you build something powerful, you also have to think carefully about how that power can be misused. News headline reading "Flock will mandate new tools to curb policy abuse of license-plate readers" over an image of a pole-mounted surveillance camera ## Why This Matters I think this is something founders often underestimate. When you're building a startup, it's natural to focus almost entirely on the product. You want to ship faster, talk to customers, improve retention, and find product-market fit. Security, fraud, abuse, privacy, compliance and infrastructure can feel like problems for later. Until they're suddenly not. A vulnerability doesn't care whether you're still a 10-person startup. A fraudulent customer doesn't care how early your company is. And a payment failure doesn't become less frustrating just because you're still figuring out product-market fit. That's why I think founders should treat infrastructure as part of the product from much earlier on. You don't necessarily need to build every piece yourself. You need to make sure the important pieces are being handled. That's one of the reasons we built Dodo Payments the way we did. A founder should be able to integrate payments quickly, go live without spending months building financial infrastructure, and then keep improving the product while the complicated parts of payments run in the background. The goal isn't to give founders another system to manage. It's to remove systems they shouldn't have to build in the first place. ## What We Shipped That philosophy also shows up in our latest release, [Dodo Payments v1.112.0](https://docs.dodopayments.com/changelog/v1.112.0). We're constantly refining the infrastructure so founders can spend less time maintaining payments and more time building their businesses. We rebuilt the [Webhooks](https://docs.dodopayments.com/developer-resources/webhooks) experience directly inside the dashboard, making it easier to create endpoints, inspect delivery attempts, replay failed events, test integrations, and monitor errors. You can also configure email alerts when webhook deliveries start failing, without changing your existing integration. We also made [Discount Codes](https://docs.dodopayments.com/features/discount-codes) significantly more flexible. You can now use fixed-amount discounts, schedule campaigns, define per-currency options, and control whether a code is available to everyone, first-time customers, existing customers, or specific users. Alongside those, we shipped: - **[Cash App Pay](https://docs.dodopayments.com/changelog/v1.112.0#3-cash-app-pay-for-subscriptions)** for subscriptions, bringing recurring payments to US USD checkouts. - **[SEPA Direct Debit](https://docs.dodopayments.com/changelog/v1.112.0#4-sepa-direct-debit)** for one-time EUR payments across the Eurozone. - **[Clearer payment failure messages](https://docs.dodopayments.com/changelog/v1.112.0#5-clearer-payment-failure-messages)**, giving customers understandable explanations while giving merchants actionable error information. - **[Analytics refinements](https://docs.dodopayments.com/changelog/v1.112.0#9-analytics-new-widgets-and-refinements)**, including a new revenue-by-country map, improved trend charts, persistent filters, new date ranges, and other dashboard improvements. None of these updates fundamentally change what payments are. They make the existing experience easier to build on, easier to operate, and more adaptable to where payments are going next. Dodo Payments discount code editor showing a flat 500 INR amount discount with per-currency options for INR and USD ## One Last Thought The more capable technology becomes, the more important the guardrails around it become. AI is making it easier to build products, automate workflows, and move faster than ever. But it is also making some forms of abuse faster and more scalable. The same is true for payments, data, and almost every other piece of infrastructure a startup depends on. You don't need to build everything yourself. You just need to make sure the things you depend on are strong enough to let you keep building. Build fast. But build with the guardrails to keep going. Also, join our loving [Discord community](https://discord.gg/dodo-payments-1305511580854779984)! Best, **Rishabh Goel** **Co-Founder, Dodo Payments** --- - [More Newsletter articles](https://dodopayments.com/blogs/category/newsletter) - [All articles](https://dodopayments.com/blogs)