Models & Labs · United States

Steelian Blocks AI Agent-Based Web Attacks in In-House Tests

Credited to TheElec · thelec.net · 2h ago

Steelian announced on Oct. 6 that it successfully blocked attacks in its in-house tests where artificial intelligence agents identified website APIs and automatically operated browsers. The company also stated that it defended against credential-stuffing attacks, which use stolen account credentials to gain unauthorized access to other services.

The tests utilized WebSuit, Steelian's AI-based web security solution. The system blocked both API reconnaissance using the command-line tool cURL and access via the browser automation tool Playwright. Steelian noted that these tests were modeled after actual attacks recently observed in the financial sector.

WebSuit randomizes the API addresses used during communications between websites and servers, while encrypting the data exchanged between them. These measures are designed to prevent attackers from inspecting communications and mimicking requests. The server accepts a request only after verifying that it contains a valid authentication token and that the data remains unaltered.

To counter AI agent attacks, Steelian applied AI behavior analysis and proof-of-work (PoW) challenges. The system analyzes user behavior—including mouse movements, clicks, inputs, and scrolling—alongside the browser execution environment to distinguish human users from automated programs. Suspicious requests are required to solve computational problems before being accepted, an approach intended to increase the time and resources required by attackers.

WebSuit mitigates credential stuffing through token verification and automation detection technologies. Depending on the customer's environment, these technologies can be applied in stages ranging from monitoring to verification and blocking, according to the company.

"To effectively respond to AI agent attacks, we must verify defensive measures at every stage, from exploring the service structure to executing requests," said Kim Byung-chul, head of Steelian's solutions business division. "Based on research and validation from an attacker's perspective, we will continue to advance web and API protection technologies that customers can readily apply to real-world services."

Models & LabsUnited States

Kore.ai launches Autoloop to keep tuning enterprise AI agents after they go live

Credited to SiliconANGLE · siliconangle.com4h ago

Useful

Enterprise artificial intelligence platform company Kore.ai Inc. today launched Autoloop, an optimization engine for the AI agents that customers build on its Kore.ai Agent Platform.

Businesses set the targets, and Autoloop keeps adjusting the agents to hit them automatically, including after they’re deployed.

Autoloop goes after the way most enterprises maintain agents today, fixing failures by hand one at a time. Patching one problem often opens up another, according to Kore.ai. In the company’s “ 2026 Kore.ai Agent Productivity Index, ” 79% of enterprises reported reversing an action taken by an AI agent. A failure their teams could not trace had hit 70% of them.

Customers set up Autoloop by giving it goals covering measures such as task completion, adherence to business rules and cost, and the engine optimizes for all of them at once. Accuracy is one of the goals as well, judged on whether answers are backed by the enterprise’s own data. Every proposed change is scored against the full set, so lower token spending cannot quietly come at the cost of safety or the agent’s ability to finish the job.

Autoloop writes the first version of each agent, tests included, from the operating procedures a company already has, then keeps iterating until every goal is met. Real interactions start fresh rounds of optimization once the agent is in production.

Spotting where a goal slipped falls to Kore.ai’s StateTrace evaluation layer. Agents are judged against the full record of what they did in production, down to each handoff, tool call and state change across a network of agents. A five-layer validation architecture makes most of the checks deterministic. The company says that keeps round-the-clock optimization affordable at enterprise scale.

Kore.ai’s Agent Blueprint Language handles the fix. Introduced earlier this year, the language compiles routing, business rules and guardrails into an executable state machine, so each step in a trace maps back to a specific part of the blueprint. Autoloop can then rewrite only the piece responsible for a miss.

“You can’t optimize what you can’t see, or fix precisely what you can’t express precisely,” said Prasanna Arikala, Kore.ai’s chief technology officer and chief product officer. He credited the two technologies with making automatic optimization practical.

The approach behind Autoloop also shapes Kore.ai’s own software development, where AI agents write code. About 6,500 commits a month to the company’s 2.6 million-line production codebase now come from those agents, which work under 68 always-on guardrails. Founder and Chief Executive Raj Koneru said the companies that manage to scale AI “will be the ones using AI to build, govern and optimize AI.”

Autoloop is available now to all customers on the Kore.ai Agent Platform’s Artemis edition, which launched in May. Kore.ai counts more than 500 Global 2000 organizations as customers.

Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.

Are you an AWS customer? Support SiliconANGLE financially by buying your AWS services from our Marketplace portal page and links: https://siliconangle.com/aws-marketplace/

Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.

OneSpan Sign takes aim at the slow work around bank agreements

Stuut cashes in on agentic order-to-cash automation with $52.5M in funding

Atlassian lays groundwork for humans and AI agents to work side by side

Caterpillar and CoreWeave shorten the learning loop for physical AI

Google goes nuclear to power its AI data centers, committing to buy 890 megawatts of energy from Constellation

Meta teams up with Bret Taylor's Sierra Technologies on new standards for AI agent commerce

Join our community

Original · SiliconANGLE

Front Method

Models & Labs · United States

Why staff using ChatGPT, Claude and Gemini for work could create problems for employers

Credited to The National · thenationalnews.com · 5h ago

Photograph · The National

Cyber security experts have urged companies to be vigilant against the risks posed by employees using outside AI applications including Claude, ChatGPT and Gemini in the workplace.

The practice, known as using shadow AI, can lead to important data being leaked and, in some cases, malware and viruses being uploaded into companies' internal systems.

Firms are urged to ensure they are aware of what AI tools are being used by employees when it comes to sensitive company information, as it could lead to problems with customer information and intellectual property.

Pasting documents into an outside AI application, or even asking it to tidy up an email, means that sensitive information is being sent to a cloud tool outside the company. If the person's personal accounts are hacked, their company's data could end up in the wrong hands.

“Staff might see the benefits of completing work faster, but employers are likely to have no record of the systems they were using or the information those systems received,” said Vladislav Tushkanov, research and development group manager at the cyber security company Kaspersky. “You cannot send personally identifiable information in the cloud.”

This means employees will need to be educated on the need to ensure that only internal AI systems are used to process information, added Mr Tushkanov, who spoke to The National at the Ai Everything summit taking place this week in Abu Dhabi.

The rapid advancements in AI mean it is crucial that companies take a fresh look at business workflows and ensure security is built in from the start, said another expert, Premchand Kurup, chief executive of the cyber security firm ParamountAssure.

“Companies should first identify unapproved AI use that could expose customer information or intellectual property through external tools, then establish approved alternatives and clear access limits,” Mr Kurup said.

The scale of the risks companies face from data breaches was laid out in IBM’s 2026 Middle East Cost of a Data Breach Report, which found 26 per cent of malicious breaches were AI-enabled. The average breach cost in the region was $8 million. The research included organisations in the UAE and Saudi Arabia.

Monitoring the AI agents

Another challenge firms will face is keeping a closer eye on AI agents that play a more advanced role as technology improves. Agentic AI is likely to carry out more tasks with less human supervision, raising questions of who is accountable when something goes wrong.

“We can't put AI agents in jail, so ultimately it will always have to be a human [who is accountable],” said Costi Perricos, Deloitte’s global generative AI leader.

Businesses need to manage agents as a digital workforce, with performance checks and decisions about how much responsibility each system is given, said Mr Perricos.

Their behaviour could change over time as they pursue their goals, he said. Companies need to track what agents are doing, establish when they are ready for use and know when to switch them off.

An example of how quickly AI capability is increasing is provided by the UAE government, which has set a target to introduce agentic AI across 50 per cent of federal sectors and operations within two years, under a framework announced in April.

Companies should set access limits and require AI-assisted security investigations to draw on verified internal records, said Mr Kurup. He added that analysts need evidence they can check before approving sensitive actions.

“The same discipline should guide wider AI adoption, with employees trained to challenge outputs and records kept for review,” he said.

Ahmed Hafez, director of solution engineering for the Middle East, Turkey and Africa at the data company Snowflake, said businesses need trusted information and controls over what AI agents can access.

“Extensive testing in controlled environments, with governance and security built in from the ground up, becomes increasingly important as these systems take on more complex work,” he said.