Security and AI news from the week beginning 4 May 2026

There has been a lot of talk about the AI bubble. Much of that has focused on the forward orders for infrastructure made by AI vendors, with limited revenue to substantiate those orders. This week, we saw some interesting numbers that suggest we are far from a bubble.

NVIDIA announced deals with Corning ($3.2 bilion) and IREN ($2.1 billion) this week. Corning will increase its U.S.-based optical connectivity manufacturing capacity by 10x and expand its U.S. fiber production capacity by more than 50% to meet the accelerating demand driven by AI factory buildouts. The expansion includes the construction of three new advanced manufacturing facilities in North Carolina and Texas.

The deal with IREN will see the two companies support the deployment of up to 5 gigawatts of NVIDIA DSX-aligned AI infrastructure across IREN’s global data centre pipeline over time. The initial focus is on the deployment of NVIDIA accelerated compute in DSX AI factories to expand access to AI-native, startup and enterprise customers.

Anthropic has reported explosive growth for its first quarter of 2026. Annualised revenues exploded from $9 billion to over $30 billion. The company had projected growth of around 10%, but these numbers show an 80-fold increase in usage and revenue in a single quarter. With the company still expected to push for an IPO, this would value it at $1.2 trillion.

While the US Govt has been outspoken about the EU AI Act, it seems it is considering its own set of controls. National Economic Council Director Kevin Hassett dropped hints on a policy shift on TV this week. The expectation is that this will be an executive order rather than support for legislation, and the process is likely to be similar to FDA drug approvals, where models undergo safety testing before public release.

BlackFog’s latest report shows that the rate of data exfiltration remains critically high at 96%. Qilin was the most active variant, responsible for 22 attacks (8%).  ShinyHunters followed with 16 attacks (6%), and INC accounted for 11 attacks (4%). Notably, 38% of all publicly disclosed ransomware incidents were not attributed to any known group.

ShinyHunters stole 275 million records from Canvas. It has had a worldwide impact across the education sectors, and the group is threatening to release messages if its demands are not met. While no financial data was lost, names, emails, and course details were taken.

A security researcher has revealed that Microsoft Edge exposes all your passwords in clear text. If a machine is compromised, then this would give attackers access to those credentials. A Microsoft spokesperson said that this is by design. “Browsers access password data in memory to help users sign in quickly and securely — this is an expected feature of the application” Not so much Security by Design, but credential leak by design.

In Case You Missed It

Pinecone has released Pinecone Nexus, a knowledge engine designed to move reasoning from retrieval to curation. It will stop agents from burning tokens on raw data and start giving them trusted, compiled knowledge. Agents are burning tokens because they are working with systems designed for humans. It causes them to retrieve, read, discard, and, because of missing data, repeat again and again, until eventually they have the answer.

In a podcast, Bennett Indart, Senior Vice President of Product Incubation at NTT Research, talked about Scale Academy. He talks about how he plans to turn research into products and deliver an ROI on the R&D that NTT undertakes. The first product will be Salt Grain, an attribute-based encryption (ABE) product.

Sachin Agrawal, Managing Director of Zoho UK, talks on this podcast about compliance being a benefit and not just friction. That is a view not shared by a lot of people who just see compliance as increasing their costs.

FourKites has released significant enhancements to Inventory Twin, the inventory intelligence layer within the FourKites Intelligent Control Tower. FourKites says the new capabilities close the disconnect between supply chain planning and execution. This is one of the most persistent gaps in enterprise supply chains.

Aera has announced new agentic reasoning capabilities in its decision intelligence solution. Agents can now identify the right data, in the right context and take the correct action. It moves the use of agents beyond the surfacing of insights and into a new phase. Importantly, this isn’t about sidelining the human-in-the-loop. They still have approval control over any actions that are taken.

WSO2 is addressing the problem of agent sprawl with the release of WSO2 Agent Manager. The speed at which enterprises are racing to deploy AI agents is causing problems. They can’t track agents, they don’t know what privileges they have, and agents are becoming more prevalent than humans on the network. From a security perspective, it’s the wild west. There is a lack of control, visibility and governance over agents.

Precisely has added new capabilities to its Data Integrity Suite to help organisations make better use of agentic-ready data. This is the most integrated, governed and enriched data that they hold. It is designed to help agentic agents find the data and context they require to answer the questions asked of them. The Precisely Data Integration Assistant is its latest AI tool to help users find those answers.

Bluevoyant has appointed John Hernandez as its next CEO. He takes over the role from co-founder James Rosenthal, who will serve as the Chairman of the company’s Board of Directors. Rosenthal has overseen the growth of Bluevoyant to become a leader in Managed Detection and Response (MDR), Third-Party Risk Management (TPRM), and Digital Risk Protection (DRP), and cyber defence-related professional services.

In Other News

Datadog

Datadog, Inc. has achieved FedRAMP (Federal Risk and Authorization Management Program) High certification, which signifies Datadog’s ability to meet one of the federal government’s most stringent cloud security and compliance standards.

eSentire

Atlas now triages every user-reported phishing email with AI — automatically. From the moment an employee clicks Report Phish to a structured verdict and recommended response action, the entire workflow is automated, auditable, and tenant-wide. User Reported Phishing is a new Atlas capability that replaces manual phishing triage with an end-to-end AI-powered pipeline.

Qualys

Qualys, Inc. and Converge have announced a joint offering that rewards organisations for demonstrated cybersecurity compliance. The collaboration allows Qualys customers who actively manage and prove strong security hygiene with Enterprise TruRisk Management (ETM) to potentially qualify for reduced cyber insurance premiums from Converge.

Securonix

Securonix, Inc. announced the Securonix Threat Research Agent and ThreatWatch for ThreatQ, expanding how security teams research threats, validate exposure, and turn intelligence into documented action. It is built on the ThreatQ platform and connected to Securonix security operations workflows.

Quantexa

Quantexa has announced the opening of its new Dublin office at 5 George’s Dock. Located in the heart of Dublin’s International Financial Services Centre (IFSC), it marks a major expansion of the company’s European research and development footprint in Europe.

Security and AI news from the week beginning 27 April 2026

The post Security and AI news from the week beginning 4 May 2026 appeared first on Enterprise Times.