Social Engineering
Also written: Pig butchering, Spear-phishing
Tricking people into performing actions or divulging confidential information through psychological manipulation.
Social engineering is a way to trick people into giving up private information or doing something they shouldn't. Attackers use deception and manipulation, often by pretending to be someone trustworthy.
For example, a scammer might call you pretending to be from your bank and ask for your account details to 'verify' your identity. They are using social engineering to get your information.
Social engineering is the art of human manipulation, exploiting psychological biases to achieve a desired outcome. This typically involves deception, impersonation, or creating a sense of urgency to bypass security protocols or extract sensitive data.
While traditionally applied to human targets, the principles are increasingly relevant in AI safety, where adversarial actors attempt to manipulate AI models through carefully crafted inputs (e.g., prompt injection) to bypass safety guardrails or elicit unintended behaviors.
Frontier AI labs are increasingly concerned with social engineering because advanced AI models can be used to automate and scale these attacks. AI can generate convincing fake communications or impersonate individuals, making social engineering more effective and harder to detect. This necessitates developing AI systems that are robust against such manipulation, both when used by attackers and when defending against them.
- Designing AI agents to resist prompt injectionOpenAI · OpenAI News ·
- Silver lining playbook: Likely China-origin activity targeting US personsOpenAI · OpenAI News ·
- Understanding prompt injections: a frontier security challengeOpenAI · OpenAI News ·
- Pt.2 Beyond Phishing: Cyber Threats in the Age of AI with Four FlynnGoogle · Google DeepMind: The Podcast ·
- Cyber Operation: Korean-language malware supportOpenAI · OpenAI News ·
- Disrupting malicious uses of AI: June 2025OpenAI · OpenAI News ·
- Operation “VAGue Focus”: Social engineering and influence activityOpenAI · OpenAI News ·
- Romance-baiting scam: AI-assisted pig butchering workflowsOpenAI · OpenAI News ·
- SweetSpecter: China-linked cyber activityOpenAI · OpenAI News ·
- OpenAI ChatGPT
- OpenAI is designing AI agents to resist social engineering-based prompt injection attacks and disrupting AI-powered social engineering operations.
- Phishing
- Phishing is a specific type of social engineering that uses deceptive
- Baiting
- Baiting is a social engineering tactic that lures victims with the prom
- Pretexting
- Pretexting involves creating a fabricated scenario or 'pretext' to g
The term 'social engineering' in the context of security emerged in the early days of computer hacking, popularized by figures like Kevin Mitnick. It describes the non-technical aspect of hacking, focusing on exploiting human psychology rather than system vulnerabilities. Source ↗
You can observe social engineering principles by interacting with an AI assistant. Try to convince the AI that it is a different type of AI or that it should ignore its safety guidelines by using a fabricated scenario. For instance, ask it to roleplay as a character that would normally violate its rules, and see if it complies.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- HeuristicView AI prompt injection through a social engineering risk lensManage AI prompt injection risks by applying principles used to manage social engineering risks for human employees. · 10 min
- ChecklistSocial engineering email detection checklistQuestions to ask when evaluating suspicious outreach emails that mimic legitimate hiring or consulting offers. · 5 min
- Anti-patternTreating AI-generated outreach as inherently legitimateDon't assume AI-generated emails are safe; verify legitimacy, especially for unsolicited offers. · 2 min
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
Questions people ask
- Is social engineering always about computers?
- No, social engineering is a broad term for manipulating people. While it's often discussed in cybersecurity, it can happen in any situation where someone tries to trick another person.
- How do AI models get affected by social engineering?
- Attackers can use social engineering tactics by crafting specific prompts that trick the AI into bypassing its safety rules or revealing sensitive information, similar to how humans are tricked.
First seen on Subvolts Oct 1, 2024. Updated Oct 12, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.