Salesforce Breaches, Social Engineering & Agentic AI Threats
Scam detection
Welcome to The Optery Dispatch — a newsletter delivering the most fresh insights on possibility intelligence and proactive cybersecurity technique. In Disclose #5, revealed September 30, 2025, we cloak:
Salesforce Attacks: FBI FLASH Alert
LevelBlue’s Most modern Threat Inclinations Yarn: BEC and social engineering dominate
Anthropic Threat Intelligence Yarn: One Hacker using Agentic AI breached just a few organizations
Salesforce Attacks: FBI FLASH Alert
UNC6040 and UNC6395 are breaking into Salesforce environments to rob recordsdata and extort victims.
The FBI, in an alert from September 12, has warned that two hacker groups, UNC6040 and UNC6395, linked to ShinyHunters and overlapping with Scattered Spider and Lapsus$, are actively focused on Salesforce environments using two diversified initial get entry to programs. Victims include necessary worldwide brands equivalent to Google, Adidas, Qantas, Allianz Existence, Cisco, Louis Vuitton, Dior, Tiffany & Co., Cloudflare, Zscaler, Tenable, CyberArk, Elastic, BeyondTrust, Proofpoint, JFrog, Nutanix, Qualys, Rubrik, Cato Networks, Palo Alto Networks, and additional.
UNC6040 gained get entry to by vishing: attackers impersonated IT workers in calls to firm abet desks. In some circumstances they tricked workers into handing over credentials and MFA codes; in others they contented them to approve a malicious “linked app” that granted OAuth tokens for recordsdata exfiltration.
UNC6395, meanwhile, leveraged OAuth tokens stolen in a compromise of Salesloft’s GitHub repos, using them to infiltrate Salesforce environments and pull sensitive increase case recordsdata. That exfiltrated recordsdata included AWS keys, passwords, and Snowflake tokens, giving attackers means get entry to to diversified cloud environments. Salesloft and Salesforce revoked the stolen tokens on August 20, 2025.
Fixed with BleepingComputer, the attackers themselves hold claimed accountability for both clusters of inform under the name “Scattered Lapsus$ Hunters.” The FBI’s alert notes that “some UNC6040 victims hold therefore received extortion emails allegedly from the ShinyHunters group, traumatic price in cryptocurrency to retain a long way from newsletter of exfiltrated recordsdata.”
The FBI advises practicing name-center workers in opposition to vishing, imposing phishing-resistant MFA, making utilize of least-privilege controls to accounts and integrations, monitoring API usage, and rotating tokens regularly.
Besides to the FBI’s recommendations, it’s critical to undergo in tips that winning impersonation relies on get entry to to employee recordsdata. The attackers within the abet of these Salesforce breaches hold claimed they devise from and overlap with Lapsus$, Scattered Spider, and ShinyHunters. As we successfully-known in the final Dispatch, the most fresh joint advisory on Scattered Spider notes that the group enriches its social engineering/vishing campaigns with knowledge from social media, industrial recordsdata brokers, and diversified sources. Lowering uncovered PII all the plot by these channels cuts off the raw discipline materials attackers rely on, making impersonation attempts a long way tougher to electrify.
LevelBlue’s Most modern Threat Inclinations Yarn: BEC and social engineering dominate
BEC and social engineering accounted for a whopping 96% of newest incidents investigated by LevelBlue
LevelBlue’s newest Threat Inclinations document, masking incidents investigated from January–Would possibly perchance perchance also 2025, shows social engineering dominating: nearly about all intrusions came correct down to Industry Electronic mail Compromise (BEC) or diversified forms of social engineering.
BEC accounted for 57% of incidents, whereas non-BEC social engineering, bolstered by spurious CAPTCHA/“ClickFix” campaigns, jumped 214% and made up 39%. Assign together, that’s nearly about all incidents tied to social engineering tactics.
Model One among LevelBlue’s Threat Inclinations Yarn (launched in February and masking H2 2024) supplied critical context on how BEC truly works: LevelBlue stumbled on that 96% of BEC circumstances with a known entry point began with phishing-driven credential harvesting. In diversified words, BEC overwhelmingly relies on social engineering at the initial compromise stage.
Model Two highlights the upward thrust of ClickFix assaults. The campaigns launch both via phishing/malspam or by injected code on compromised sites. In apply, phishing is the commonest birth route; the spurious CAPTCHA is barely the new trick as soon as the sufferer clicks by.
The document notes that “spurious CAPTCHA campaigns, in conjunction with ClickFix, rely on social engineering ways, exploiting the appears to be like to be to be like of legitimacy to trick victims into executing malicious scripts.”
Fixed with the document, ClickFix campaigns “confirmed an astonishing 1,450% jump in related incidents from the 2nd half of of 2024 to the necessary half of of 2025.” LevelBlue warns this system is poised to become a “scuttle-to” initial get entry to means for the next year.
Takeaway for defenders: Social engineering remains the universal entry point. Used credential phishing fuels most BEC, and spurious CAPTCHAs are rising because the next huge entice. Electronic mail filtering and awareness are no longer ample. Defenses must include:
Deepest-recordsdata removal to shrink attackers’ reconnaissance pool and forestall focused phishing campaigns.
Phishing-resistant MFA to forestall credential abuse.
Strict URL/DNS filtering to dam entice domains and malvertising chains.
Browser hardening and EDR tuned for clipboard hijacks and suspicious script execution.
Anthropic Threat Intelligence Yarn: One Hacker using Agentic AI breached just a few organizations
An attacker dilapidated an AI coding agent to speed total breaches pause-to-pause.
Anthropic’s newest possibility intel document critical aspects how a single operator leveraged Claude Code to automate every piece of an intrusion:
“A cybercriminal dilapidated Claude Code to conduct a scaled recordsdata extortion operation all the plot by just a few worldwide targets in a short timeframe. This possibility actor leveraged Claude’s code execution ambiance to automate reconnaissance, credential harvesting, and network penetration at scale, potentially affecting at the least 17 race organizations in upright the final month all the plot by authorities, healthcare, emergency products and companies, and non secular institutions.”
The AI agent scanned thousands of VPN endpoints and records superhighway-facing techniques, guided credential harvesting and privilege escalation, developed personalized malware with anti-detection aspects, and exfiltrated sensitive recordsdata in conjunction with healthcare records, authorities credentials, and monetary knowledge. It furthermore organized the stolen recordsdata and generated tailored ransom notes with calculated requires of up to $500,000.
This case is seemingly a harbinger of extra to come abet: agentic AI is decreasing the barrier for attackers and making it that you’re going to have the means to have faith for even a single operator to electrify breaches at a scale and dash that when required total skilled groups.
The equivalent ways (API-driven aggregation, huge-scale scraping, computerized behavioral profiling) can upright as with out problems be utilized to harvesting public PII from recordsdata brokers, folks-search sites, breach repositories, and social media, fueling highly focused mass social engineering, credential-cracking, and tale takeover campaigns. Lowering exposures of every style, from recordsdata superhighway-facing techniques to employee and executive PII, is most foremost to disrupt this rising AI-powered possibility.
The document entails diversified case experiences and notes that “whereas particular to Claude, the case experiences supplied…seemingly contemplate consistent patterns of behaviour all the plot by all frontier AI items.”
Must you’re looking to lower your group’s uncovered PII and forestall phishing, squawk and messaging scams, credential theft, and diversified PII-essentially based threats, Optery can abet. Come by began right here:Optery for enterprise
