AMD Denies $10,000 Bug Bounty for RCE Flaw Despite Researcher’s Cooperation AMD has refused to pay a $10,000 bug bounty to security researcher Paul, who reported a remote code execution (RCE) vulnerability in the company’s auto-updater software. The flaw, discovered via a man-in-the-middle (MITM) attack, was initially rejected under AMD’s bug bounty policy, which excluded MITM-based exploits. Despite the denial, Paul cooperated with AMD, temporarily removing his public disclosure at the company’s request. AMD promised to issue a CVE, fix the software, and credit Paul but no bounty. The researcher later regretted the agreement, as AMD repeatedly extended the disclosure timeline, citing delays in patching "multiple tools" beyond the original scope. The fix, released on June 9 124 days after the initial report involved replacing an insecure "http" link with "https" in the code. While AMD overhauled the updater’s download mechanism, Paul confirmed the new version now securely fetches drivers. However, the software still relies on the outdated CRC32 hash for file validation, which lacks cryptographic security. Adding irony to the situation, a Reddit user later noted the vulnerable code wasn’t even active, meaning the updater was already broken preventing AMD from pushing fixes until users manually downloaded the patched version. The case highlights ongoing tensions between researchers and vendors over bug bounty policies, particularly when critical vulnerabilities are downplayed or excluded from payouts.

AMD disclosed a critical security flaw named RMPocalypse (CVE-2025-0033) in its Secure Encrypted Virtualization with Secure Nested Paging (SEV-SNP) mechanism, affecting multiple EPYC processor series (7003, 8004, 9004, 9005, and Embedded variants). The vulnerability stems from incomplete protections in the Reverse Map Paging (RMP) table initialization, allowing attackers with admin-level hypervisor access to exploit a race condition during AMD Secure Processor (PSP) setup.Exploitation enables arbitrary memory corruption, bypassing SEV-SNP’s confidentiality and integrity guarantees. Attackers can inject malicious code, forge security attestations, replay old states, or activate debug modes, leading to full compromise of confidential virtual machines (CVMs) and 100% success rate in exfiltrating secrets. While no evidence of active exploitation exists, the flaw undermines cloud security foundations, particularly in Azure Confidential Computing (ACC) and enterprise environments relying on AMD’s hardware-based isolation.Patches are available for most affected processors, though Embedded 7003 and 9005 series fixes are delayed until November 2025. The vulnerability highlights systemic risks in trusted execution environments (TEEs), where initialization gaps can nullify all subsequent security assurances.

The ransomware group ShinyHunters (Scattered Lapsus$ Hunters) breached Salesforce by exploiting stolen OAuth tokens from Salesloft Drift’s AI chatbot integration, compromising 1.5 billion records across 760 companies (including Cisco, Disney, and Marriott). The leaked data includes PII (names, DOBs, passports, employment histories), shipping details, chat transcripts, flight records, and car ownership data—validated by cybersecurity researchers. Attackers first infiltrated Salesloft’s GitHub repository, extracting private source code and OAuth tokens, then laterally moved to Google Workspace, Microsoft 365, and Okta platforms of victims. The group demanded separate ransoms from Salesforce and listed 39 high-profile victims on a darkweb leak site, pressuring them to pay under threat of full data exposure. The attack leveraged social engineering (vishing, phishing, IT impersonation) to trick employees into granting access, highlighting vulnerabilities in third-party supply-chain integrations and weak 2FA/OAuth security controls.

Researchers discovered a critical vulnerability in AMD's Zen CPUs, termed 'EntrySign,' which allows attackers with high privileges to install malicious microcode by exploiting the AES-CMAC algorithm's flaw used in validation processes. This vulnerability affects AMD Zen architecture CPUs from versions 1 to 4, enabling attackers to bypass cryptographic checks and potentially gain persistent access to manipulate the processors' instruction set. The impact of such an attack could be especially severe if the compromised CPUs are used in cloud services and AI infrastructures, posing risks to data integrity, system reliability, and the security posture of affected entities.

AMD disclosed a security vulnerability, designated as CVE-2024-36347 with a CVSS score of 6.4, affecting a broad range of processors. This flaw, discovered by Google researchers, lies in the improper signature verification of AMD’s CPU ROM microcode patch loader, allowing attackers with administrative privileges to load unauthorized microcode patches. The vulnerability has serious implications for system integrity and confidentiality, potentially resulting in compromised execution, data breaches, and System Management Mode (SMM) environment threats. While no real-world attacks have been reported, the theoretical impact could be severe, necessitating timely firmware updates to mitigate risks.

RansomHouse gang claimed to have stolen 450 GB of data from the semiconductor giant AMD in a recent cyber attack. The stolen data from the firm includes research and financial information, which they were analyzed to determine its value after adding it to their data leak site. The compromised data includes a leaked a CSV containing a list of over 70,000 devices that appear to belong to AMD's internal network, as well as an alleged list of AMD corporate credentials for users with weak passwords

The disclosure of the Sinkclose vulnerability, affecting AMD processors since 2006, showcases a significant security oversight allowing hackers to gain access to the privileged System Management Mode. By implanting a bootkit, attackers gain persistent, undetectable control over a system, monitoring activity and surviving system reinstalls. The flaw exposes countless systems to a level of compromise where the only solution may be the physical disposal of the infected machine. This vulnerability not only undermines the trust in device security but also signifies immense potential losses in data integrity and financial repercussions for both AMD and affected users.