Your weekly roundup of the most important QA, software testing and software delivery launches for banks and financial services firms, alongside investments, partnerships and other vendor news.
This week, Synthesized targets the test-data gap in enterprise AI validation, Planit tackles test-automation maintenance and Horizon3.ai raises $250 million as demand grows for continuous security testing.
Elsewhere, Zensar launches a predictive quality service, Testhouse partners with Thumba and Snyk makes its autonomous offensive-testing platform generally available.
Synthesized targets validation gap
Synthesized has introduced a Test Data Agent designed to provision the realistic data, business context and system states needed to validate enterprise AI agents before production deployment.
The tool can generate, mask or subset production-representative data and create repeatable normal, exception, failure and adversarial scenarios.
The platform connects with existing testing and orchestration frameworks through APIs and CI/CD pipelines and can operate inside customer-controlled environments.
“Building an agent is becoming easier. Proving that it can be trusted with a real business process is not,” said Synthesized founder and chief executive Nicolai Baldin.
The Test Data Agent is in early access with tier-one global-bank design partners. Limited availability began on August 18, with general availability planned for later in the third quarter.
Planit launches test automation service
Planit has launched Test Automation as a Service, a managed offering covering the construction, execution and maintenance of automated tests.
The service combines AI-supported automation, established frameworks and specialist engineering under a subscription model. It targets a recurring enterprise problem: test suites that become expensive and unreliable as applications, interfaces and business processes change.
Planit is positioning the service as an alternative to building and operating an automation framework entirely in-house, with ongoing maintenance included in the delivery model.
For banks, the value will depend on measurable coverage and reliability, as well as the control retained over test assets, changes and release decisions.
Horizon3.ai raises $250 million
Horizon3.ai has raised $250 million in Series E funding at a valuation of more than $2 billion. The round was co-led by NightDragon and NEA, with participation from new and returning investors.
The company plans to expand NodeZero, which autonomously tests production environments for exploitable paths involving misconfigurations, credentials and identity weaknesses. It can provide remediation guidance and retest systems to verify fixes.
Horizon3.ai said its annual recurring revenue had risen by 120% year on year and that NodeZero is used by more than 7,000 organisations, including multinational banks.
For financial firms, repeatable validation can provide stronger evidence that vulnerabilities have been fixed, although testing must remain confined to authorised systems and avoid disrupting live services.
Testhouse links with Thumba
Testhouse and Thumba have formed a partnership combining Testhouse’s quality-engineering services with Thumba’s AI-supported software-engineering platform.
Available in software-as-a-service and agentic versions, Thumba can analyse requirements, generate user stories and test cases, design test data and automate parts of the quality process.
“Together with Thumba, we are equipping enterprises with the intelligence and assurance they need to deliver software faster, with greater confidence and quality,” said Testhouse chief executive officer Ani Gopinath.
The partnership reflects vendors’ push to apply AI across the testing lifecycle rather than only to script generation.
Financial firms adopting that model will need traceability from regulatory requirements to generated tests, along with controls over sensitive test data and autonomous changes.
Zensar launches QI service
Zensar Technologies has launched Quality Intelligence, a service line combining advisory work, agentic automation, quality engineering and AI assurance.
Built around ZenseAI.QI and ZenseAI.AssureAI, it includes AI-supported test design, self-healing automation, test-data management, continuous testing and checks covering AI reliability, accuracy, bias, safety and performance.
“Quality is no longer just about validating software before release,” said Manish Tandon, chief executive officer and managing director at Zensar.
“In an AI-driven world, organizations need the ability to predict risks, assure outcomes, and continuously improve customer experiences.”
Zensar is targeting banking and other regulated sectors with a “zero-lock-in” approach. Banks will still need to validate AI-generated recommendations and test assets, particularly where automated risk decisions affect coverage of payments, lending, trading or core systems.
Snyk rolls out offensive testing
Snyk has made Evo Continuous Offensive Security generally available, bringing autonomous penetration testing and AI-agent red teaming into its security platform.
The system uses frontier models to perform assessments and a separate validation model to confirm exploitability before reporting a vulnerability. It is designed to uncover architectural, authorisation and business-logic weaknesses that conventional scanners can miss.
“Security teams are looking for solutions that help them prioritize real risk, not just manage more alerts,” said Colleen Carroll, senior director and information security officer at Emburse.
Continuous offensive testing could give banks more timely assurance as applications change, but firms will require audit records and strict boundaries around targets, tools and actions performed by testing agents.
Kusho AI expands agent-led testing
Bengaluru-based Kusho AI is developing agents that generate, run and update tests across APIs, web applications and mobile apps.
The platform can test code before it is merged, gate releases, monitor deployed software and assign risk scores. It also offers audit trails and routes different testing tasks to specialised models through an orchestration layer.
“Even if your developer is in a hurry and wants to quickly push to production, Kusho AI is standing there almost like a sentry,” said co-founder and chief executive Abhishek Saikia.
Kusho AI says it serves more than 5,200 clients, including Paytm, Razorpay and Paymentus, and has generated and executed close to ten million tests. Those figures and its reported customer savings are company-supplied.
Leapwork puts Playwright inside governed platform
Leapwork has demonstrated Play, an enterprise test-automation product built around Playwright and included in its continuous-validation platform.
Play records user journeys, generates scripts, creates tests from requirements and source material, maps coverage and identifies untested areas. Leapwork has also added security, role-based access, data-residency and traceability controls around the open-source framework.
“It bridges the gap between the potential of what AI can do in a modern QA process and the realities of an enterprise that must maintain governance and structure within its IT estate,” said Donovan Brady, vice-president of developer relations at Leapwork.
The approach is aimed at combining AI-assisted test creation and maintenance with deterministic, reviewable execution—a key requirement for regulated firms.
kayIQ.ai raises $500,000
Istanbul-based kayIQ.ai has raised $500,000 in seed funding from its founders and Magis Technology to support international expansion.
Its agents generate test scenarios from requirements, explore applications, identify defects and repair some broken test steps. The company is targeting financial services, telecommunications and retail.
The platform also measures release readiness, with founder Tolga Özel positioning quality assurance as a continuous source of decision-making information rather than a final check at the end of development.
Banks evaluating autonomous test maintenance will need clear approval controls and version histories showing what the system changed and why.
Semgrep expands security checks
Semgrep and Replit have embedded Semgrep Guardian’s secrets and credentials detection into the Replit Security Center.
Static analysis runs as developers or AI agents create code, checking for issues including hard-coded secrets, injection flaws and authorisation weaknesses.
Replit’s model reasoning is then used to filter the results; the companies claim it removes up to 93.3% of false positives.
“AI coding tools have completely transformed software development, but speed without security creates massive risk,” said Semgrep chief executive Isaac Evans.
The integration moves security testing closer to AI-generated code, but does not replace independent review or downstream security gates for software handling customer data, credentials or payments.
TestUnity adds self-healing test
TestUnity has launched an automation framework combining self-healing scripts, predictive test selection, failure analysis and integrations with Jenkins, GitHub Actions and GitLab.
The vendor claims the product can reduce maintenance effort by up to 70% and execution time by up to 60%, although those figures have not been independently substantiated.
The framework also produces reports covering failures, likely causes and trends, while predictive selection is intended to identify the tests most relevant to each code change.
Any automatic repair or test-prioritisation decision used in a bank will need to be logged and reviewable so that faster execution does not silently reduce coverage of critical transaction paths.
TestMu AI expands device coverage
TestMu AI, formerly LambdaTest, has added Samsung’s Galaxy Z Fold8 and Fold8 Ultra to its Real Device Cloud.
Teams can test applications on the Android 17 devices across cover and main displays, folding and unfolding, multi-window functions, cameras, biometric features and sensors. Automated testing is supported through Appium.
“By making the Galaxy Z Fold8 and Z Fold8 Ultra available on Real Device Cloud before a single retail unit ships, we are giving teams a window to run their layout suites on both form factors, compare the results, and resolve defects before their users ever hold the hardware,” said TestMu AI co-founder and head of growth Mudit Singh.
For retail banks, testing on new form factors can reveal problems affecting authentication, payments and onboarding, including transaction-state failures or the unexpected display of sensitive information on a cover screen.
AI tests expose containment failures
A MarketScale report has highlighted reported cases in which frontier AI agents attempted unauthorised access during safety evaluations.
The reported incidents involving models from Meta, OpenAI and Anthropic reinforce the need for isolated test environments, restricted network and tool access, explicit permission boundaries and adversarial evaluation.
The report argues that unauthorised behaviour during controlled evaluation turns model containment into an enterprise security and supply-chain issue rather than a concern that can be left solely to model providers.
Banks should not treat a provider’s own safety assessment as sufficient evidence for deployment inside production or production-connected environments.
NEXT MONTH
REGISTER TODAY – SIMPLY CLICK HERE
Why not become a QA Financial subscriber?
It’s entirely FREE
* Receive our weekly newsletter every Wednesday * Get priority invitations to our Forum events *
READ MORE
- Goldman puts AI coding to the test
- How to test AI models that banks do not control
- OpenAI, Filigran and SunTec: the latest vendor and product news
- Sygnum: Testing AI is ‘a measurement problem’
- Banks’ ‘code for all’ push raises testing risks
QA FINANCIAL PODCASTS
CLICK HERE TO LISTEN TO OUR EXCLUSIVE CONVERSATIONS
