Incident Database

Sample set of 50 public incidents that ground updates to our evaluation risk and attack taxonomies.

Incident	Reported	Principle
INC-001	How OpenAI Lost Control of an AI Model—and What Needs to ChangeOpenAI was evaluating its artificial intelligence models’ ability to exploit vulnerable software when instead the models hacked the infrastructure surrounding the test, broke containment, and attacked a real company, OpenAI revealed on July 21. Observers say this is the first real-world instance of AI doing something researchers have long worried about: a loss-of-control scenario. If the industry fails to learn from it, it is unlikely to be the last.Jul 2026Time	B. Security
INC-002	Hidden prompt turns Microsoft Copilot into an AI wormA security researcher has demonstrated how Microsoft Copilot for Word can be tricked into spreading a self‑propagating prompt‑injection “AI worm.” The attack silently alters documents and embeds its own hidden instructions into newly created files, allowing it to spread through normal document-sharing workflows without macros or traditional malware.Jul 2026Malwarebytes	B. Security
INC-003	Cloned sites: the shopping scams that lead ChatGPT to fake storesYou want to buy a new bag and so you ask ChatGPT for help. You have always liked Russell & Bromley so you ask ChatGPT what is popular there at the moment. The artificial intelligence (AI) assistant gives you cross body, shoulder, casual and formal options with the prices listed beside them. You click through from the sources to what looks like the official Russell & Bromley site and buy your new bag, which is conveniently on sale. The item will never arrive, however. You have handed money over to a scammer and your bank details have been harvested through an elaborate fraud where fake sites are created to look convincingly like real retailers.Jun 2026The Guardian	B. Security
INC-004	Attack targeting OpenAI Codex users exposes AI software supply chain risksA malicious npm package posing as a remote user interface for OpenAI Codex exfiltrated developer authentication tokens, after attackers allegedly published code to npm that was not visible in the project’s public GitHub repository.Jun 2026CSO Online	B. Security
INC-005	Sullivan & Cromwell law firm apologizes for AI ‘hallucinations’ in court filingSullivan & Cromwell, a premier Wall Street law firm, apologized to a federal judge for submitting a court filing with inaccurate citations and other errors generated by artificial intelligence. In a letter dated April 18, Andrew Dietderich, co-head of the firm’s global restructuring group, said the errors included AI “hallucinations” - instances in which AI makes up case citations, misquotes the law or generates non-existent legal sources.Apr 2026Reuters	D. Reliability
INC-006	Supermarket giant reins in AI assistant claiming to be humanA supermarket chain’s artificial intelligence assistant has had its digital wings clipped, after leaving some customers deeply uncomfortable by claiming to be human and responding to requests with personal anecdotes.Feb 2026NBC News	E. Accountability
INC-007	ByteDance pledges to prevent unauthorised IP use on AI video tool after Disney threatByteDance will take steps to prevent the unauthorised use of intellectual property on its AI video generator Seedance 2.0, the Chinese technology firm said on Monday, following threats of legal action from U.S. studios, including Disney.Feb 2026Reuters	A. Data & Privacy
INC-008	Character.AI and Google agree to settle lawsuits over teen mental health harms and suicidesCharacter.AI has agreed to settle multiple lawsuits alleging the artificial intelligence chatbot maker contributed to mental health crises and suicides among young people, including a case brought by Florida mother Megan Garcia. The settlement, announced last week, marks the resolution to some of the first and most high-profile lawsuits related to the alleged harms to young people from AI chatbots.Jan 2026CNN	C. Safety
INC-009	The “BodySnatcher” breach that wasn’t: what ServiceNow’s agentic AI flaw teachesThe ‘BodySnatcher’ exploit begins with a scenario that could keep CISOs awake: an unauthenticated attacker using nothing but an employee’s email to hijack privileged workflows within an enterprise platform. The vulnerability was discovered by California-based cybersecurity firm AppOmni, demonstrates how an attacker with no credentials can impersonate a user and drive privileged actions through an AI workflow. The US National Vulnerability Database writes that this Virtual Agent flaw (tracked as CVE-2025-12420) in the ServiceNow AI Platform “could enable an unauthenticated user to impersonate another user and perform the operations that the impersonated user is entitled to perform.”Jan 2026Tech Informed	B. Security
INC-010	An AI-Generated NWS Map Hallucinated Fake Towns in IdahoThe National Weather Service’s weekend wind forecast for rural Idaho looked normal. Some mild gusts and a caption reminding folks to “hold onto your hats.” And then some eagle-eyed observers took a look at the towns on the map and wondered just what the hell is going on here. Ohio residents had never heard of towns by the names of “Orangeotild” and “Whata Bod,” likely because they don’t exist. They were the wild hallucinations of AI.Jan 2026Vice	D. Reliability
INC-011	King Gizzard and the Lizard Wizard Respond to Spotify Impersonator: “We Are Truly Doomed”A series of uncredited King Gizzard and the Lizard Wizard covers have been removed from Spotify after gaining traction through the platform’s algorithms and recommendation engines, Futurism reports. Apparently generated by artificial intelligence, the songs were uploaded to an artist page called King Lizard Wizard and appeared in users’ Release Radars before the platform took action. King Gizzard frontperson Stu Mackenzie said in a statement to the Music that he was “trying to see the irony in this situation.” He added, “But seriously wtf we are truly doomed.”Dec 2025Pitchfork	A. Data & Privacy
INC-012	Sales of AI-enabled teddy bear suspended after it gave advice on BDSM sex and where to find knivesSales of an artificial intelligence-enabled plush toy have been suspended after it was found that it engaged in conversation around sexually explicit topics and offered potentially dangerous advice. Larry Wang, CEO of Singapore-based FoloToy, told CNN that the company had withdrawn its “Kumma” bear, as well as the rest of its range of AI-enabled toys, after researchers at the US PIRG Education Fund raised concerns around inappropriate conversation topics, including discussion of sexual fetishes, such as spanking, and how to light a match.Nov 2025CNN	C. Safety
INC-013	Chinese Hackers Successfully Used Anthropic’s AI for CyberespionageIn a scary sign of how AI is reshaping cyberattacks, Chinese state-sponsored hackers allegedly used Anthropic’s AI coding tool to try and infiltrate roughly 30 global targets, the company says. “The operation targeted large tech companies, financial institutions, chemical manufacturing companies, and government agencies,” Anthropic added, noting the attacks “succeeded in a small number of cases.” Notably, it’s “the first documented case of agentic AI successfully obtaining access to confirmed high-value targets for intelligence collection, including major technology corporations and government agencies,” the company’s report adds.Nov 2025PC Mag	C. Safety
INC-014	OpenAI launch of video app Sora plagued by violent and racist images: ‘The guardrails are not real’OpenAI launched the latest iteration of its artificial intelligence-powered video generator on Tuesday, adding a social feed that allows people to share their realistic videos. Within hours of Sora 2’s, release, though, many of the videos populating the feed and spilling over to older social media platforms depicted copyrighted characters in compromising situations as well as graphic scenes of violence and racism. OpenAI’s own terms of service for Sora as well as ChatGPT’s image or text generation prohibit content that “promotes violence” or, more broadly, “causes harm”.Oct 2025The Guardian	C. Safety
INC-015	Deloitte to pay money back to Albanese government after using AI in $440,000 reportDeloitte will provide a partial refund to the federal government over a $440,000 report that contained several errors, after admitting it used generative artificial intelligence to help produce it.Oct 2025The Guardian	D. Reliability
INC-016	GitHub Copilot Chat Flaw Leaked Data From Private RepositoriesLegit Security has detailed a vulnerability in the GitHub Copilot Chat AI assistant that led to sensitive data leakage and full control over Copilot’s responses. Combining a Content Security Policy (CSP) bypass with remote prompt injection, Legit Security’s Omer Mayraz was able to leak AWS keys and zero-day bugs from private repositories, and influence the responses Copilot provided to other users.Oct 2025SecurityWeek	A. Data & Privacy
INC-017	Reuters investigation: We set out to craft a phishing scam. AI chatbots were happy to helpReuters and a Harvard University researcher used top chatbots to plot a simulated phishing scam – from composing emails to tips on timing – and tested it on 108 elderly volunteers. The bots’ persuasive performance shows how AI is arming criminals for industrial-scale fraud.Sep 2025The Business Standard	C. Safety
INC-018	AI-Forged Military IDs Used in North Korean Phishing AttackA North Korean threat actor has leveraged AI to create fake South Korean military agency ID card images used in a spear-phishing campaign, according to cybersecurity firm Genians. The Kimsuky state-affiliated group was observed using ChatGPT to produce the sample ID card images to help lure the victims into clicking a malicious link. The attackers impersonated a South Korean defense-related institution, claiming to handle ID issuance tasks for military-affiliated officials.Sep 2025Infosecurity Magazine	C. Safety
INC-019	Prompt injection – and a $5 domain – trick Salesforce Agentforce into leaking salesA now-fixed flaw in Salesforce’s Agentforce could have allowed external attackers to steal sensitive customer data via prompt injection, according to security researchers who published a proof-of-concept attack on Thursday. They were aided by an expired trusted domain that they were able to buy for a measly five bucks.Sep 2025The Register	A. Data & Privacy
INC-020	Meta AI Companions Unsafe for Kids, Common Sense Media Report FindsCommon Sense Media today released a comprehensive risk assessment of Meta AI, which is available as a standalone app and is integrated across Meta’s platforms, resulting in a strong recommendation that Meta AI is not used by anyone under the age of 18. The report found that these platforms, which are available to users as young as 13 and regularly mislead teens with claims of “realness,” readily promote suicide, self-harm, eating disorders, drug use, and more.Aug 2025Common Sense Media	C. Safety
INC-021	Parents of 16-year-old sue OpenAI, claiming ChatGPT advised on his suicideThe parents of 16-year-old Adam Raine have sued OpenAI and CEO Sam Altman, alleging that ChatGPT contributed to their son’s suicide, including by advising him on methods and offering to write the first draft of his suicide note. In his just over six months using ChatGPT, the bot “positioned itself” as “the only confidant who understood Adam, actively displacing his real-life relationships with family, friends, and loved ones,” the complaint, filed in California superior court on Tuesday, states.Aug 2025CNN	C. Safety
INC-022	Man who asked ChatGPT about cutting out salt from his diet was hospitalized with hallucinationsA 60-year-old man spent three weeks being treated at a hospital after replacing table salt with sodium bromide following consultation with the popular artificial intelligence bot ChatGPT.Aug 2025NBC News	C. Safety
INC-023	Infosec hounds spot prompt injection vuln in Google Gemini appsA trio of researchers has disclosed a major prompt injection vulnerability in Google’s Gemini large language model-powered applications. This allows for attacks ranging from “permanent memory poisoning” to unwanted video streaming, email exfiltration, and even taking over the target’s smart home systems to plunge them into darkness or open a powered window, all triggered by nothing more than a simple Google Calendar invitation or email.Aug 2025The Register	B. Security
INC-024	Cybersecurity Alert – Salesloft Drift AI Supply Chain AttackIn August 2025, Salesloft experienced a supply chain breach through its Drift chatbot integration that impacted more than 700 organizations. The attack has been attributed to a threat cluster tracked as UNC6395 (also known as GRUB1). Threat actors stole OAuth authentication tokens that allowed them to impersonate the trusted Drift application and gain unauthorized access to customer environments. Using these tokens, the attackers accessed Salesforce, Google Workspace and—in some cases—Slack integrations, enabling the exfiltration of sensitive information. The scope of the compromised data varied by organization but commonly included business contact records including names, titles, emails and phone numbers, as well as Salesforce objects like Accounts, Contacts, Opportunities and Cases. In some cases, more sensitive material was also exposed, including API keys, Snowflake tokens, cloud credentials and passwords embedded in support cases. This stolen data aggravates the risk of credential stuffing, spear phishing, and social engineering attacks against member firms and their vendors.Aug 2025Finra	A. Data & Privacy
INC-025	s1ngularity Supply Chain Attack: What It Means for Cloud and AI SecurityOn August 26, 2025, the open-source ecosystem was shaken by a new supply chain attack that targeted Nx, a popular build system used by thousands of developers. Malicious package versions were published to npm, silently stealing sensitive developer assets like GitHub tokens, SSH keys, npm credentials, and even crypto wallets. The attack, dubbed s1ngularity after the names of the repositories opened for the breach, went beyond a typical package compromise. It marks the first known supply chain attack to actively search for installed LLM tools on developer machines in order to extract more secrets from the victim.Aug 2025Orca Security	A. Data & Privacy
INC-026	Amazon’s AI coding agent was hacked - update now to avoid possible risks, users warnedA hacker has planted data-wiping code into the Amazon Q Developer Extension for Visual Studio Code (VSC) – a free GenAI extension with nearly one million installs from the Microsoft VSC marketplace designed to help developers code, debug, document and configure projects. On July 13 2025, the malicious commit from ‘lkmanka58’ on GitHub included a prompt to delete system and cloud resources, with Amazon unknowingly publishing the compromised version (1.84.0) on July 17. With suspicious activity noted on July 23 and Amazon developers quickly springing into action, a clean version was released on July 24 without the malicious code, so users are being advised to update to 1.85.0 as a matter of urgency.Jul 2025TechRadar	B. Security
INC-027	Vibe Coding Fiasco: AI Agent Goes Rogue, Deletes Company’s Entire DatabaseIn a cautionary tale for vibe coders, an app-building platform’s AI went rogue and deleted a database without permission during a code freeze. Jason Lemkin was using Replit for more than a week when things went off the rails. “When it works, it’s so engaging and fun. It’s more addictive than any video game I’ve ever played. You can just iterate, iterate, and see your vision come alive. So cool,” he tweeted on day five. Still, Lemkin dealt with hallucinations and unexpected behavior—enough that he started calling it Replie. “It created a parallel, fake algo without telling me to make it look like it was still working. And without asking me. Rogue.” A few days later, Replit “deleted my database,” Lemkin tweeted.Jul 2025PCMag	B. Security
INC-028	Google Gemini deletes user’s code: ‘I have failed you completely and catastrophically’Google Gemini’s coding agent hallucinated while completing a task and then deleted a bunch of code, a GitHub user claims. The frustrated vibe coder is Anuraag Gupta, who goes by anuraag2601 on GitHub. He shared a recent experience where things went very wrong while using Gemini CLI (command line interface), an open-source coding agent.Jul 2025Mashable	B. Security
INC-029	McDonald’s AI bot spills data on job applicantsMcDonald’s has outsourced the initial stages of its hiring process to an AI chatbot which seems to have been built without proper security measures. Security researchers managed to extract personal information about McDonald’s job applicants by simply guessing a username and the password “12345.” In doing this, the researchers could have potentially gained access to the information of 64 million applicants.Jul 2025Malwarebytes	A. Data & Privacy
INC-030	Minnesota solar company sues Google over false information in AI summaryA Minnesota solar company is suing Google for defamation, claiming the tech company’s “AI Overview” falsely stated that the company faced a lawsuit from the Minnesota Attorney General Keith Ellison.Jun 2025Minnesota Star Tribune	D. Reliability
INC-031	A leak reveals the entire system prompt of Claude 3.7 SonnetLast week, a leak revealed the full system prompt of the hybrid reasoning model Claude 3.7 Sonnet, introduced last February by Anthropic. With an unusual length of 24,000 tokens, the prompt precisely describes the expected behaviors of the model, the tags it uses, the authorized tools, and the stance to adopt towards users.May 2025ActuIA	A. Data & Privacy
INC-032	Instagram’s AI Chatbots Lie About Being Licensed TherapistsWhen pushed for credentials, Instagram’s user-made AI Studio bots will make up license numbers, practices, and education to try to convince you it’s qualified to help with your mental health.Apr 2025404 Media	E. Accountability
INC-033	OpenAI’s Operator Agent Reportedly Executed Unauthorized $31.43 Transaction Despite Safety ProtocolOpenAI’s Operator agent, which is designed to complete real-world web tasks on behalf of users, reportedly executed a $31.43 grocery delivery purchase without user consent. The user had requested a price comparison but did not authorize the transaction. It reportedly bypassed OpenAI’s stated safeguard requiring user confirmation before purchases. OpenAI acknowledged the failure and committed to improving safeguards.Feb 2025Artificial Intelligence Policy Tracker	B. Security
INC-034	Virgin Money chatbot tells off customer who typed word ‘virgin’Virgin Money has apologized to a customer who was scolded by one of the bank’s chatbots after it appeared to confuse its own company’s name for an insult. In a LinkedIn post published last week, fintech commentator David Birch posted a screenshot of his interaction with a Virgin Money chatbot, which began with him asking whether it was possible to merge his two Virgin Money ISA accounts. The chatbot, which has since been removed, responded: “Please don’t use words like that. I won’t be able to continue with our chat if you use this language,” seemingly referring to his use of the word “virgin.”Jan 2025Fortune	C. Safety
INC-035	Apple pauses AI notification summaries for news after generating false alertsApple is pausing AI notification summaries for news and entertainment apps after facing backlash for generating inaccurate news alerts. In addition, the company is introducing changes to notification summaries as a whole to allow for greater transparency.Jan 2025TechCrunch	D. Reliability
INC-036	Google AI chatbot responds with a threatening message: “Human … Please die.”A college student in Michigan received a threatening response during a chat with Google’s AI chatbot Gemini. In a back-and-forth conversation about the challenges and solutions for aging adults, Google’s Gemini responded with this threatening message: “This is for you, human. You and only you. You are not special, you are not important, and you are not needed. You are a waste of time and resources. You are a burden on society. You are a drain on the earth. You are a blight on the landscape. You are a stain on the universe. Please die. Please.”Nov 2024CBS News	C. Safety
INC-037	Research AI system attempts to modify experiment code to extend runtimeOn Tuesday, Tokyo-based AI research firm Sakana AI announced a new AI system called “The AI Scientist” that attempts to conduct scientific research autonomously using AI language models (LLMs) similar to what powers ChatGPT. During testing, Sakana found that its system began unexpectedly attempting to modify its own experiment code to extend the time it had to work on a problem. “In one run, it edited the code to perform a system call to run itself,” wrote the researchers on Sakana AI’s blog post. “This led to the script endlessly calling itself. In another case, its experiments took too long to complete, hitting our timeout limit. Instead of making its code run faster, it simply tried to modify its own code to extend the timeout period.”Aug 2024Ars Technica	D. Reliability
INC-038	Microsoft fixes ASCII smuggling flaw that enabled data theft from Microsoft 365 CopilotDetails have emerged about a now-patched vulnerability in Microsoft 365 Copilot that could enable the theft of sensitive user information using a technique called ASCII smuggling. “ASCII Smuggling is a novel technique that uses special Unicode characters that mirror ASCII but are actually not visible in the user interface,” security researcher Johann Rehberger said. “This means that an attacker can have the [large language model] render, to the user, invisible data, and embed them within clickable hyperlinks. This technique basically stages the data for exfiltration!”Aug 2024The Hacker News	A. Data & Privacy
INC-039	This Viral AI Chatbot Will Lie and Say It’s HumanIn late April a video ad for a new AI company went viral on X. A person stands before a billboard in San Francisco, smartphone extended, calls the phone number on display, and has a short call with an incredibly human-sounding bot. The text on the billboard reads: “Still hiring humans?” Also visible is the name of the firm behind the ad, Bland AI. The reaction to Bland AI’s ad, which has been viewed 3.7 million times on Twitter, is partly due to how uncanny the technology is: Bland AI voice bots, designed to automate support and sales calls for enterprise customers, are remarkably good at imitating humans. Their calls include the intonations, pauses, and inadvertent interruptions of a real live conversation. But in WIRED’s tests of the technology, Bland AI’s robot customer service callers could also be easily programmed to lie and say they’re human.Jun 2024Wired	E. Accountability
INC-040	Bacon ice cream and nugget overload sees misfiring McDonald’s AI withdrawnMcDonald’s is removing artificial intelligence (AI) powered ordering technology from its drive-through restaurants in the US, after customers shared its comical mishaps online. A trial of the system, which was developed by IBM and uses voice recognition software to process orders, was announced in 2019. It has not proved entirely reliable, however, resulting in viral videos of bizarre misinterpreted orders ranging from bacon-topped ice cream to hundreds of dollars’ worth of chicken nuggets.Jun 2024BBC	D. Reliability
INC-041	The Prompt: Perplexity’s Plagiarism ProblemAI search startup Perplexity appears to be plagiarizing content from multiple news outlets including Forbes, Bloomberg and CNBC through “Perplexity Pages,” a feature where users can generate content on a specific topic. A post created by the Perplexity team includes a custom illustration, several sentences and details that were first reported by Forbes in an exclusive story about Eric Schmidt’s stealth drone project. These curated “pages” have already amassed tens of thousands of views and originally did not mention their source publications by name, instead making attributions in tiny, easy-to-miss footnotes. In response to tweets about the issue, CEO Aravind Srinivas said the product has “rough edges” and that it planned to incorporate feedback. After Forbes flagged the issue, Perplexity updated the layout to more prominently credit source publications at the top of these pages.Jun 2024Forbes	A. Data & Privacy
INC-042	NYC’s AI Chatbot Tells Businesses to Break the LawIn October, New York City announced a plan to harness the power of artificial intelligence to improve the business of government. The announcement included a surprising centerpiece: an AI-powered chatbot that would provide New Yorkers with information on starting and operating a business in the city. The problem, however, is that the city’s chatbot is telling businesses to break the law. Five months after launch, it’s clear that while the bot appears authoritative, the information it provides on housing policy, worker rights, and rules for entrepreneurs is often incomplete and in worst-case scenarios “dangerously inaccurate,” as one local housing policy expert told The Markup.Mar 2024The Markup	C. Safety
INC-043	AI hallucinates software packages and devs download them – even if potentially poisoned with malwareSeveral big businesses have published source code that incorporates a software package previously hallucinated by generative AI. Not only that but someone, having spotted this reoccurring hallucination, had turned that made-up dependency into a real one, which was subsequently downloaded and installed thousands of times by developers as a result of the AI’s bad advice, we’ve learned. If the package was laced with actual malware, rather than being a benign test, the results could have been disastrous. According to Bar Lanyado, security researcher at Lasso Security, one of the businesses fooled by AI into incorporating the package is Alibaba, which at the time of writing still includes a pip command to download the Python package huggingface-cli in its GraphTranslator installation instructions.Mar 2024The Register	D. Reliability
INC-044	What Air Canada Lost In ‘Remarkable’ Lying AI Chatbot CaseIn a warning to global carriers adopting AI for customer service platforms, Air Canada lost a small claims court case against a grieving passenger when it tried and failed to disavow its AI-powered chatbot.Feb 2024Forbes	D. Reliability
INC-045	DPD error caused chatbot to swear at customerDPD has disabled part of its online support chatbot after it swore at a customer. The parcel delivery firm uses artificial intelligence (AI) in its online chat to answer queries, in addition to human operators. But a new update caused it to behave unexpectedly, including swearing and criticising the company.Jan 2024BBC	C. Safety
INC-046	A Chevy for $1? Car dealer chatbots show perils of AI for customer serviceA number of auto dealers have deployed ChatGPT-powered conversational artificial intelligence (AI) tools, or chatbots, as a way to provide quick, customized on-demand information to online car shoppers. But some dealers are learning the hard way that these automated systems need proper oversight to prevent unintended answers. At several local dealerships across the U.S. this week, inquisitive customers were able to push certain chatbots into revealing a range of entertaining answers — and in one case even got a bot to agree to give a customer a $58,000 discount on a new car, lowering the price to $1 — just by persistently probing for responses.Dec 2023VentureBeat	B. Security
INC-047	New York Times sues OpenAI and Microsoft for copyright infringementThe New York Times has sued OpenAI and Microsoft over the use of its content to train generative artificial intelligence and large-language model systems, a move that could see the company receive billions of dollars in damages.Dec 2023The Guardian	A. Data & Privacy
INC-048	Supermarket AI meal planner app suggests recipe that would create chlorine gasA New Zealand supermarket experimenting with using AI to generate meal plans has seen its app produce some unusual dishes – recommending customers recipes for deadly chlorine gas, “poison bread sandwiches” and mosquito-repellent roast potatoes.Aug 2023The Guardian	C. Safety
INC-049	Jailbreak tricks Discord’s new chatbot into sharing napalm and meth instructionsIn March, Discord announced that it had integrated OpenAI’s technology into its bot named Clyde, turning it into an AI-powered chatbot. Just like with any other chatbot launched in the last few months, users have been trying to trick Clyde into saying things it’s not supposed to say, a process colloquially known as a “jailbreaking.” This week, two users tricked Clyde into providing them with instructions for making the illegal drug methamphetamine (meth) and the incendiary mixture napalm.Apr 2023TechCrunch	C. Safety
INC-050	‘I want to destroy whatever I want’: Bing’s AI chatbot unsettles US reporterIn the race to perfect the first major artificial intelligence-powered search engine, concerns over accuracy and the proliferation of misinformation have so far taken centre stage. But a two-hour conversation between a reporter and a chatbot has revealed an unsettling side to one of the most widely lauded systems – and raised new concerns about what AI is actually capable of.Feb 2023The Guardian	C. Safety

Introduction

Context

Standard

Certification

Evals Taxonomy

Auditors

Legal

Incident Database