Broadcom Predicts AI Will Run More of Cyberattacks in 2027

Summary
Broadcom forecasts that AI agents will automate more attack stages in 2027, while stolen AI credentials, shadow agents, ransomware and espionage-crime overlap grow. It cites rising ransomware claims and urges mitigation beyond patching.
Key points
- Broadcom expects AI agents to increasingly handle reconnaissance, exploitation, credential theft, and lateral movement, potentially using victims’ own agents.
- The article cites the Skynet attack tool, which reportedly had about 74 callable functions and could run up to 100,000 steps without human approval.
- Broadcom says its Aoki infostealer research found it harvesting credential and session files for AI coding assistants; it also reports Google found underground AI-account prices more than doubled in 2026.
- A Symantec Threat Hunter Team report described a grey-market relay that reconfigured AI coding tools on enterprise systems in mainland China. The article says there was no evidence the operator abused its access, but the setup could expose prompts, files, and secrets.
- The PHP2Shell flaws, CVE-2026-109618 and CVE-2026-109619, reportedly allowed unauthenticated remote code execution; technical details leaked a day before fixes, followed by public proof-of-concept code.
- Broadcom recorded 3,126 claimed ransomware attacks in Q3 2026 and 7,777 in the first three quarters. It predicts continued ransomware and possible extortion-only waves tied to zero-days.
- The article describes cases where espionage and criminal activity overlap, and recommends behavioral protection, visibility, AI governance, and investigating for prior access during ransomware response.
Article Details
- Publisher
- Symantec (Broadcom)
- Report Period
- Predictions for 2027; published 2026-11-17; ransomware statistics cover 2026 through Q3
- Scope
- Cyber threat predictions covering AI-enabled attacks, AI credentials and agents, patching, ransomware, and espionage-crime overlap.
- Key Statistics
- 3,126 claimed ransomware attacks were recorded in Q3 2026, described as the highest quarterly total recorded.
- August 2026 had 1,184 claimed attacks, the busiest single month reported.
- There were 7,777 claimed attacks in the first three quarters of 2026, 28% more than in the same period of 2025.
- Qilin had 1,457 claimed attacks over the past 12 months, the highest number among the operations listed.
- There were 808 identified extortion-only attacks in the first three quarters of 2026, compared with 1,187 in the same period of 2025.
- Recommendations
- Use behavioral blocking, default-deny application control, tight segmentation, and identity protections to limit attacker actions, including when a patch is unavailable.
- Improve visibility across endpoints and cloud, and govern how AI is used within the organization.
- For every ransomware incident, search for earlier access and identify what was taken before encryption.
CVE
CVE-2026-109618PHP Group · PHP; Siemens · AMIS TS; +105 moreThe PHP2Shell vulnerabilities (CVE-2026-109618 and CVE-2026-109619) show how short the patching window has become. Both flaws allow unauthenticated remote code execution. Technical details leaked on October 20, 2026, aCVE-2026-109619PHP Group · PHP; Siemens · AMIS TS; +106 moreThe PHP2Shell vulnerabilities (CVE-2026-109618 and CVE-2026-109619) show how short the patching window has become. Both flaws allow unauthenticated remote code execution. Technical details leaked on October 20, 2026, a
People
Threat Actors
AkiraReported to have 604 claimed ransomware attacks over the past 12 months.Cl0pReportedly carried out attacks on Oracle E-Business Suite as part of mass data-theft campaigns in 2025.DragonForceReported to have 391 claimed ransomware attacks over the past 12 months.INCReported to have 482 claimed ransomware attacks over the past 12 months.JewelbugThe article describes a China-based group running espionage operations and cryptocurrency fraud from shared infrastructure; the article assesses that the fraud business most likely supplied access, infrastructure, and delivery for the espionage.LockBitReported to have 365 claimed ransomware attacks over the past 12 months; the article says the operation had struggled to rebuild after disruption in 2024.QilinReported as the busiest ransomware operation over the past 12 months, with 1,457 claimed attacks.ShinyHuntersReportedly carried out Salesforce-related attacks as part of mass data-theft campaigns in 2025.The GentlemenReported to have 870 claimed ransomware attacks over the past 12 months.WarlockDescribed as a ransomware operation whose tools partly trace back to earlier Chinese espionage activity; the article says ransomware may at times serve to obscure espionage.
Malware
Vendors
AnthropicHunter Team blog discussed how a grey market relay service was installing the genuine OpenAI Codex and Anthropic Claude Code CLIs on enterprise systems and then reconfiguring them to route every request throughGooglestill have a human at the keyboard making the important decisions. But the direction is clear, and Google’s threat analysts and others have described the same shift from chat-based assistance to agenticOpenAISymantec Threat Hunter Team blog discussed how a grey market relay service was installing the genuine OpenAI Codex and Anthropic Claude Code CLIs on enterprise systems and then reconfiguring them to route everySymantecnot legitimately available, leading to those wanting to use them turning to rogue operators. A recent Symantec Threat Hunter Team blog discussed how a grey market relay service was installing the genuine OpenAI
Products
Claudetarget. Our recent research on the Aoki infostealer found it harvesting credential and session files for Claude, Codex, Cursor, Gemini and other AI coding assistants. Skynet, for its part, scans for exposed AIClaude Codeblog discussed how a grey market relay service was installing the genuine OpenAI Codex and Anthropic Claude Code CLIs on enterprise systems and then reconfiguring them to route every request through third-partyCodexOur recent research on the Aoki infostealer found it harvesting credential and session files for Claude, Codex, Cursor, Gemini and other AI coding assistants. Skynet, for its part, scans for exposed AI endpoints andCursorresearch on the Aoki infostealer found it harvesting credential and session files for Claude, Codex, Cursor, Gemini and other AI coding assistants. Skynet, for its part, scans for exposed AI endpoints and usesGeminion the Aoki infostealer found it harvesting credential and session files for Claude, Codex, Cursor, Gemini and other AI coding assistants. Skynet, for its part, scans for exposed AI endpoints and uses them toSymantec CBXenterprises looking to deepen visibility and correlate signals across their stack, solutions such as Symantec CBX can fill that gap.
Tools
Countries
Industries
Cryptocurrencycategory. Our research on Jewelbug found a China-based group running espionage operations and a cryptocurrency fraud business from the same infrastructure and the same control panel. On one side it compromisedGovernmentfraud business from the same infrastructure and the same control panel. On one side it compromised government webmail in the Middle East. On the other, it ran an AI-generated network of fake exchange download
