security
Apr 30, 2026
By Teun
Oracle warns customers to upgrade after AI model threat
Oracle sent a customer security advisory warning that new AI models can speed up vulnerability discovery and exploitation, according to a leaked support note reviewed by SiliconANGLE. The company urged customers to upgrade database versions, apply recent release updates, and tighten network isolation as part of a new security posture.
Oracle has issued a customer security advisory warning that next-generation AI models can lower the cost of finding software flaws and chaining them into attacks, according to a leaked support note reviewed by SiliconANGLE. The note appears to be a response to Anthropic’s limited-release Mythos model and to broader advances in frontier AI systems.
SiliconANGLE reported that the advisory is one of the first public signs that major infrastructure vendors are adjusting their security guidance for a world where AI can help attackers read code, reason across systems and discover exploits faster. The article says the concern is not limited to one model or one vendor. OpenAI, Google and other major model providers could produce similar capabilities, the publication wrote.
The support note says Oracle is working with leading AI model providers to identify and remediate vulnerabilities before they can be weaponized. It also says customers on Oracle cloud database services, including Autonomous AI Database and Exadata Database Service, are in a better position than on-premises customers because those services benefit from automated patching.
According to the advisory excerpt quoted by SiliconANGLE, Oracle told customers that “these new recommendations are different from what we have recommended in the past” and that they will need to change existing update plans. Oracle also said it is not technically feasible to backport all of the new security fixes to older database versions or older release updates, which is why it is recommending newer versions.
The company is urging customers to upgrade to Oracle Database 19c or Oracle AI Database 26ai and to apply the latest quarterly Release Updates, or RUs, as quickly as possible. Oracle said it will continue to provide backports of publicly known or high-CVE vulnerabilities to older versions where feasible.
The note also points to upcoming updates scheduled for April and July 2026, which Oracle says will be the first to include security hardening informed directly by testing with advanced AI models. SiliconANGLE said Oracle has streamlined those updates to prioritize security fixes.
Oracle is also recommending Transparent Application Continuity, a feature that keeps databases available during rolling software updates. The company said strict network isolation is especially important and advised customers not to expose databases directly to the public internet.
SiliconANGLE said the alert comes as Oracle has been rolling out new agentic AI products and positioning its AI Database as a control plane for AI. The report argues that security needs to be built into that layer, especially as AI systems make vulnerability discovery faster and cheaper.
The article also said Oracle has not provided a formal statement or the advisory itself, and instead referred SiliconANGLE to a related blog post. SiliconANGLE said it found posts from Jenny Tsai-Smith, Oracle’s senior vice president of product management for Oracle Database, including one urging customers to take action and another describing resources in Oracle AI Factory.
In the article’s view, the practical message for Oracle customers is to treat the advisory as a prompt to move quickly on upgrades, review exposure by version and workload, and update security plans for database systems that may now face more automated scrutiny.
