Artificial intelligence and machine learning news
Found 426 articles
Weakly generated eMudhra keys exposed eight intermediate authorities. Chrome and Firefox blocked them as organizations review certificates and trust stores.
AI can find complex vulnerabilities quickly, but safe remediation depends on sandbox testing, environment context and tightly scoped permissions.
Attackers are turning trusted cloud tools into channels for control, data theft and disruption, putting identity and management permissions at the center of defense.
PwnEscape uses guest-controlled checksum metadata to corrupt IPv4 packet headers, exposing host memory and enabling root code execution on vulnerable systems.
The October attacks show why refrigeration security depends on contractor access controls, alerts for remote changes and plans to replace unsupported controllers.
Long-lived sensitive data and hard-to-update equipment complicate post-quantum security. Device inventories, protected gateways and vendor roadmaps can guide migration.
Three NGINX flaws rated 9.2 could enable crashes, denial of service or code execution. F5 urges customers to patch and audit configurations.
A Microsoft and Nvidia framework maps 30 risks and urges provenance checks, configuration testing, and ongoing safeguards for open-weight AI deployments.
A compromised ConnectWise credential was used to access Microsoft 365 tenants, prompting service suspensions and urgent customer-side reviews.
A newly identified flaw in OpenAI's skills repository raises concerns about supply chain security in AI applications, emphasizing the need for careful vetting.
Xbow USA utilized AI compute time worth $40K to uncover critical vulnerabilities in Chrome, leading to a $250K bug bounty from Google.
A significant vulnerability in the Linux kernel allowing guest-to-host escapes through the networking stack has been resolved, strengthening security protocols.
A recent zero-day vulnerability in BeyondTrust’s Remote Support software has raised alarms, allowing attackers to gain root access and control over user systems.
Hundreds of AI experts advocate for enhanced safety measures and an oversight body to regulate AI training and containment practices.
Research reveals that while AI aids malware evasion efforts, consistent malicious behavior still enables detection, limiting its effectiveness.
TypeSafe's Jev artificial intelligence model, designed for rapid machine-to-machine communication, is found to be more susceptible to attacks than traditional LLMs.
Cisco's Secure Email Gateway is vulnerable to exploitation, allowing attackers root access; updates are available, but vigilance is essential.
A newly identified vulnerability in Palo Alto Networks' PAN-OS requires immediate attention from federal agencies, with critical risk and root access implications.
A recent discovery reveals that a flaw in Claude Code's preview function may expose sensitive developer information, raising alarms about security practices in AI tools.
A vulnerability in Ivanti Connect Secure enabled attackers to log in without credentials. Patches are now available for affected versions.