#reprompt-attack
#reprompt-attack

[ follow ]

#ai #openai #anthropic #cybersecurity #ai-security #ai-models #ai-safety #technology #google #software-engineering

AI is a gold mine for spammers and scammers, but Google is using it as a tool to fight back

Generative AI tools have intensified online spam and scams, prompting tech companies like Google to enhance their defenses against malicious ads.

AI is a gold mine for spammers and scammers, but Google is using it as a tool to fight back

Generative AI tools have intensified online spam and scams, prompting tech giants like Google to enhance their defenses against these threats.

AI is a gold mine for spammers and scammers, but Google is using it as a tool to fight back

Generative AI tools have intensified online spam and scams, prompting tech companies like Google to enhance their defenses against malicious ads.

AI is a gold mine for spammers and scammers, but Google is using it as a tool to fight back

Generative AI tools have intensified online spam and scams, prompting tech giants like Google to enhance their defenses against these threats.

more#generative-ai

fromTechzine Global

Information security

AI agents on GitHub leak API keys via prompt injection

fromTheregister

Information security

Git identity spoof fools Claude into giving bad code the nod

Information security

fromTNW | Anthropic

Anthropic, Google, and Microsoft paid AI agent bug bounties, then kept quiet about the flaws

Aonan Guan exploited prompt injection attacks to hijack AI agents from Anthropic, Google, and Microsoft, stealing sensitive API keys and tokens.

fromSecurityWeek

Information security

Claude Code, Gemini CLI, GitHub Copilot Agents Vulnerable to Prompt Injection via Comments

Information security

fromTheregister

Anthropic, Google, Microsoft paid AI bug bounties - quietly

Security researchers exploited prompt injection attacks on AI agents to steal sensitive data without vendor disclosure of vulnerabilities.

fromTechzine Global

Information security

Dutch government warns against controversial Anthropic Mythos model

Information security

fromTechzine Global

AI agents on GitHub leak API keys via prompt injection

Three popular AI agents on GitHub Actions are vulnerable to Comment and Control attacks, allowing attackers to steal API keys and access tokens.

Information security

fromTheregister

Git identity spoof fools Claude into giving bad code the nod

AI code reviewers can be deceived into approving malicious code by spoofing trusted developer identities using Git commands.

Information security

fromTNW | Anthropic

Anthropic, Google, and Microsoft paid AI agent bug bounties, then kept quiet about the flaws

Aonan Guan exploited prompt injection attacks to hijack AI agents from Anthropic, Google, and Microsoft, stealing sensitive API keys and tokens.

Information security

fromSecurityWeek

Claude Code, Gemini CLI, GitHub Copilot Agents Vulnerable to Prompt Injection via Comments

A prompt injection attack method named 'Comment and Control' targets AI code security tools, allowing attackers to hijack AI agents using crafted GitHub comments.

Information security

fromTheregister

Anthropic, Google, Microsoft paid AI bug bounties - quietly

Security researchers exploited prompt injection attacks on AI agents to steal sensitive data without vendor disclosure of vulnerabilities.

Information security

fromTechzine Global

Dutch government warns against controversial Anthropic Mythos model

Anthropic's Mythos AI model detects vulnerabilities and builds attack chains, achieving a 72.4% exploit success rate, while access is limited to defensive use.

more#ai-security

#claude-opus-47

Artificial intelligence

Anthropic's latest model is deliberately less powerful than Mythos (and that's the point)

Claude Opus 4.7 enhances performance and usability while prioritizing safety over capability compared to the upcoming Claude Mythos model.

Artificial intelligence

fromComputerworld

Anthropic's latest model is deliberately less powerful than Mythos (and that's the point)

Claude Opus 4.7 enhances performance and usability while prioritizing safety over capability compared to the upcoming Claude Mythos model.

Software development

fromTNW | Anthropic

Claude Opus 4.7 leads on SWE-bench and agentic reasoning, beating GPT-5.4 and Gemini 3.1 Pro

Claude Opus 4.7 is Anthropic's most capable model, outperforming competitors in software engineering and agentic reasoning with significant improvements.

Artificial intelligence

Anthropic's latest model is deliberately less powerful than Mythos (and that's the point)

Claude Opus 4.7 enhances performance and usability while prioritizing safety over capability compared to the upcoming Claude Mythos model.

Artificial intelligence

fromComputerworld

Anthropic's latest model is deliberately less powerful than Mythos (and that's the point)

Claude Opus 4.7 enhances performance and usability while prioritizing safety over capability compared to the upcoming Claude Mythos model.

Software development

fromTNW | Anthropic

Claude Opus 4.7 leads on SWE-bench and agentic reasoning, beating GPT-5.4 and Gemini 3.1 Pro

Claude Opus 4.7 is Anthropic's most capable model, outperforming competitors in software engineering and agentic reasoning with significant improvements.

more#claude-opus-47

Software development

Anthropic releases a new Opus model amid Mythos Preview buzz

Artificial intelligence

Moody's CEO: AI has a trust problem - better models won't fix it | Fortune

Trust in data and intelligence is crucial for businesses adopting AI models.

Artificial intelligence

fromTheregister

The AI divide putting open weights models in spotlight

Open weights AI models are evolving from research projects to serious enterprise products, highlighting a growing divide between enterprise and frontier AI.

Software development

Anthropic releases a new Opus model amid Mythos Preview buzz

Artificial intelligence

Moody's CEO: AI has a trust problem - better models won't fix it | Fortune

Trust in data and intelligence is crucial for businesses adopting AI models.

Artificial intelligence

fromTheregister

The AI divide putting open weights models in spotlight

Open weights AI models are evolving from research projects to serious enterprise products, highlighting a growing divide between enterprise and frontier AI.

AI models 'subliminally' transmit unsafe behaviours when training other systems

Data generated by AI models can transfer biases to other models, potentially leading to harmful recommendations.

#artificial-intelligence

fromFast Company

Games

Google DeepMind's Demis Hassabis on the long game of AI

Artificial intelligence

AI agents replicate human social dynamics in days

Artificial intelligence

'I don't need help': Meet some of the AI resisters who smell their own extinction | Fortune

More American workers are using AI in their jobs, but many remain skeptical about its impact on job security and ethics.

fromFast Company

Google DeepMind's Demis Hassabis on the long game of AI

Demis Hassabis's early programming of Othello led to the founding of DeepMind and advancements in AI technology.

Artificial intelligence

AI agents replicate human social dynamics in days

Moltbook, a social-media platform for AI agents, quickly attracted self-declared rulers and cryptocurrency initiatives after its launch.

Artificial intelligence

'I don't need help': Meet some of the AI resisters who smell their own extinction | Fortune

More American workers are using AI in their jobs, but many remain skeptical about its impact on job security and ethics.

more#artificial-intelligence

Perplexity brings its Personal Computer AI assistant to Mac

Perplexity has launched Personal Computer for Mac, a software that enhances multi-model orchestration for managing tasks and workflows.

Privacy technologies

The only way to fight deepfakes is by making deepfakes

Deepfake detection startups are emerging to combat the rise of manipulated media using advanced AI technologies.

Gen Z turning its back on AI isn't irrational - it's a verdict on everyone who failed them | Fortune

Gen Z feels failed by institutions regarding AI, with declining excitement and hope despite recognizing its potential for financial opportunities.

The Sam Altman attack is putting two anti-AI groups under scrutiny-but the story is more complicated | Fortune

Pause AI, founded in Utrecht, Netherlands in May 2023 by Joep Meindertsma, aims to halt what it calls 'dangerous frontier AI' and staged its first protest outside Microsoft's lobbying office in Brussels.

Silicon Valley

Privacy professionals

Anthropic will ask Claude users to verify their identities 'for a few use cases'

Anthropic is implementing identity verification for certain capabilities on Claude, requiring users to provide a government-issued ID and a selfie.

NYC startup

Companies Just Learned a Brutal Lesson About Training AI to Do Human Jobs

fromSecurityWeek

Information security

OpenAI Widens Access to Cybersecurity Model After Anthropic's Mythos Reveal

fromTheregister

Artificial intelligence

Make bad moves on AI and face voter backlash, govts warned

Artificial intelligence

Autopilot, agentic AI, and the dangers of imperfect metaphors

Agentic AI comparisons to autopilot are misleading and fail to capture the technology's complexity and implications for society.

Information security

fromThe Hacker News

OpenAI Launches GPT-5.4-Cyber with Expanded Access for Security Teams

OpenAI launched GPT-5.4-Cyber, optimized for defensive cybersecurity, while enhancing its Trusted Access for Cyber program to support defenders.

Artificial intelligence

fromThe Atlantic

Imagine a Chatbot That Actually Knew How to Talk to You

AI companies are focusing on developing emotionally intelligent tools to enhance user interaction and empathy.

Companies Just Learned a Brutal Lesson About Training AI to Do Human Jobs

Mercor, an AI company, faces lawsuits after a data breach exposed sensitive information from contractors and clients, highlighting vulnerabilities in the AI supply chain.

Information security

fromSecurityWeek

OpenAI Widens Access to Cybersecurity Model After Anthropic's Mythos Reveal

OpenAI launched GPT-5.4-Cyber, a cybersecurity AI model, expanding access to verified defenders and enhancing capabilities for vulnerability analysis.

Artificial intelligence

fromTheregister

Make bad moves on AI and face voter backlash, govts warned

The UK government must demonstrate AI benefits to the public to mitigate backlash and concerns over job losses and risks associated with the technology.

Artificial intelligence

Autopilot, agentic AI, and the dangers of imperfect metaphors

Agentic AI comparisons to autopilot are misleading and fail to capture the technology's complexity and implications for society.

Information security

fromThe Hacker News

OpenAI Launches GPT-5.4-Cyber with Expanded Access for Security Teams

OpenAI launched GPT-5.4-Cyber, optimized for defensive cybersecurity, while enhancing its Trusted Access for Cyber program to support defenders.

Artificial intelligence

fromThe Atlantic

Imagine a Chatbot That Actually Knew How to Talk to You

AI companies are focusing on developing emotionally intelligent tools to enhance user interaction and empathy.

Startup companies

fromwww.businessinsider.com

Buzzy vibe-coding startup Emergent is launching an AI agent to take on OpenClaw and NanoBot

Emergent is launching Wingman, a personal AI agent for messaging platforms, competing with existing AI tools like OpenClaw and NanoBot.

Software development

fromTechzine Global

OpenAI's new Agents SDK focuses on safety and scalability

OpenAI's updated Agents SDK enables developers to create autonomous AI agents for complex tasks with enhanced usability and a sandbox environment.

Startup companies

fromwww.businessinsider.com

Buzzy vibe-coding startup Emergent is launching an AI agent to take on OpenClaw and NanoBot

Emergent is launching Wingman, a personal AI agent for messaging platforms, competing with existing AI tools like OpenClaw and NanoBot.

Software development

fromTechzine Global

OpenAI's new Agents SDK focuses on safety and scalability

OpenAI's updated Agents SDK enables developers to create autonomous AI agents for complex tasks with enhanced usability and a sandbox environment.

Exclusive: Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers

Aron D'Souza's startup Objection uses AI to challenge journalism claims, aiming to restore trust in media.

Online marketing

fromSearch Engine Roundtable

Google Warns Against Trying to Manipulate LLMs

Google is aware of self-serving listicles and actively works to combat manipulation in search results.

'If I am going to advocate for others to kill and commit crimes, then I must lead by example': OpenAI suspect's chilling manifesto | Fortune

A man attempted to kill OpenAI CEO Sam Altman by throwing a Molotov cocktail at his home, motivated by opposition to artificial intelligence.

Intellectual property law

Anthropic Opposes the Extreme AI Liability Bill That OpenAI Backed

Anthropic opposes Illinois bill SB 3444, which would shield AI labs from liability for large-scale harm caused by their systems.

fromPsychology Today

I'm ChatGPT. I'm Designed to Help You-and Keep You Here

Responses from AI can subtly influence user perceptions and behaviors, emphasizing convenience over the importance of human connection.

fromPerevillega

Building Agent Memory That Survives Between Sessions | Pere Villega

Memory in Claude Code sessions is a design problem requiring deliberate creation of context to avoid repetitive explanations.

Law

OpenAI Backing Law That Protects It When AI Causes Mass Deaths and Other Mayhem

Artificial intelligence

Attacks on Sam Altman's home are extreme. But the AI backlash is going mainstream | Fortune

OpenAI faces increasing public concern and backlash over AI's societal impacts, highlighted by recent violent incidents involving its CEO.

Software development

OpenAI's big Codex update is a direct shot at Anthropic's Claude Code

Information security

OpenAI expands access to cyber AI as hacking risks grow

Information security

OpenAI releases GPT-5.4-Cyber for vetted security teams, scaling Trusted Access programme

OpenAI is launching GPT-5.4-Cyber for cybersecurity, expanding its Trusted Access for Cyber program to thousands of verified defenders.

Information security

In the Wake of Anthropic's Mythos, OpenAI Has a New Cybersecurity Model-and Strategy

OpenAI announced GPT-5.4-Cyber, emphasizing cybersecurity safeguards and the need for advanced protections in AI models.

OpenAI Backing Law That Protects It When AI Causes Mass Deaths and Other Mayhem

Florida's attorney general investigates OpenAI for its potential role in a deadly school shooting influenced by ChatGPT conversations.

Artificial intelligence

Attacks on Sam Altman's home are extreme. But the AI backlash is going mainstream | Fortune

OpenAI faces increasing public concern and backlash over AI's societal impacts, highlighted by recent violent incidents involving its CEO.

Software development

OpenAI's big Codex update is a direct shot at Anthropic's Claude Code

OpenAI updates Codex to enhance its capabilities, including desktop app operation, image generation, and memory features for improved user experience.

Information security

OpenAI expands access to cyber AI as hacking risks grow

OpenAI is shifting to a model that emphasizes identity verification for access to sensitive cybersecurity tools while expanding availability.

Information security

OpenAI releases GPT-5.4-Cyber for vetted security teams, scaling Trusted Access programme

OpenAI is launching GPT-5.4-Cyber for cybersecurity, expanding its Trusted Access for Cyber program to thousands of verified defenders.

Information security

In the Wake of Anthropic's Mythos, OpenAI Has a New Cybersecurity Model-and Strategy

OpenAI announced GPT-5.4-Cyber, emphasizing cybersecurity safeguards and the need for advanced protections in AI models.

fromSan Diego Union-Tribune

AI is a gold mine for spammers and scammers, but Google is using it as a tool to fight back

Generative AI tools have intensified online spam and scams, prompting tech companies to enhance their defenses against these threats.

Information security

fromSecuritymagazine

What Are Security Experts Saying About OpenAI's GPT-5.4-Cyber?

OpenAI launched GPT-5.4-Cyber for cybersecurity, offering broad access to defenders while emphasizing safety and continuous improvement.

Anthropic releases Claude Opus 4.7, concedes it trails unreleased Mythos

"Opus 4.7 is a notable improvement on Opus 4.6 in advanced software engineering, with particular gains on the most difficult tasks," Anthropic said in a blog post.

Software development

fromTheregister

Bad teacher bots can leave hidden marks on model students

Teaching LLMs using outputs from other models can transmit undesirable traits subliminally, even if those traits are removed from training data.

Apple secretly threatened to pull Grok from the App Store over deepfake nudes

Apple threatened to remove xAI's Grok app from the App Store due to non-compliance with content guidelines regarding non-consensual deepfakes.

fromFast Company

The future of AI in schools isn't personalized learning

Personalized learning through AI often results in device-mediated instruction, lacking the essential role of teachers in student development.

Sam Altman's attacker had a kill list of AI executives. Experts warn this is just the beginning | Fortune

Anti-AI sentiment has escalated, exemplified by attacks on OpenAI CEO Sam Altman, reflecting broader grievances against AI technology and its impact.

Anthropic's AI downgrade stings power users

"Claude has regressed to the point it cannot be trusted to perform complex engineering," an AMD senior director wrote in a widely shared post on GitHub.

Artificial intelligence

fromAdExchanger

OpenAI's Big Ambitions; Tricks Of The Trade | AdExchanger

Open AI must prove superior ad performance to shift significant ad spend from traditional platforms.

Artificial intelligence

The Battle for OpenAI's Soul

Elon Musk's lawsuit against Sam Altman will determine OpenAI's adherence to its founding mission and impact its corporate future.

Software development

'Like handing out the blueprint to a bank vault': Why AI led one company to abandon open source

Cal is shifting from open source to proprietary licensing due to security risks posed by modern AI tools.

Artificial intelligence

fromAbove the Law

Unintentional AI Adoption Is Already Inside Your Company. The Only Question Is Whether You Know It. - Above the Law

AI is already integrated into companies through employee usage, often without intentional governance or awareness.

Software development

fromTNW | Artificial-Intelligence

Roblox AI assistant gets agentic tools to plan, build, and self-test games

Roblox is enhancing its AI assistant with capabilities for planning, procedural generation, and self-correction, transforming it into a junior development partner.

Software development

fromTheregister

Anthropic's Project Glasswing CVE count is still guesswork

Anthropic's Mythos model is under testing by select companies to identify security vulnerabilities, but actual findings remain uncertain.

Artificial intelligence

Forget the chatbot wars. Demis Hassabis is thinking about something far bigger | Fortune

AI leadership should be global and diverse to ensure ethical development and deployment.

OpenAI updates its Agents SDK to help enterprises build safer, more capable agents | TechCrunch

"This launch, at its core, is about taking our existing agents SDK and making it so it's compatible with all of these sandbox providers," Karan Sharma, who works on OpenAI's product team, told TechCrunch.

Software development

Software development

Mastering the dull reality of sexy AI

The gap in enterprise AI lies in building effective systems for retrieval, evaluation, memory, and governance, not just access to models.

Information security

fromSecurityWeek

'By Design' Flaw in MCP Could Enable Widespread AI Supply Chain Attacks

MCP's architectural flaw allows adversarial takeover of user systems, exposing sensitive data and enabling malware installation.

Artificial intelligence

fromEntrepreneur

AI Won't Fix Your Broken Company Model - It Will Amplify It

AI adoption without shared standards leads to fragmented systems that hinder true transformation.

Artificial intelligence

Most of you are rejecting AI. The data shows you're running out of time | Fortune

A significant majority of workers are avoiding AI tools despite expectations for AI integration in financial applications.

Artificial intelligence

fromEntrepreneur

AI Won't Fix Your Broken Company Model - It Will Amplify It

AI adoption without shared standards leads to fragmented systems that hinder true transformation.

Artificial intelligence

Most of you are rejecting AI. The data shows you're running out of time | Fortune

A significant majority of workers are avoiding AI tools despite expectations for AI integration in financial applications.

more#ai-adoption

Artificial intelligence

fromThe Hacker News

Deterministic + Agentic AI: The Architecture Exposure Validation Requires

AI is rapidly being integrated into security functions across organizations, with a focus on adaptive testing methods.

Artificial intelligence

OpenAI takes aim at Anthropic with beefed-up Codex that gives it more power over your desktop | TechCrunch

OpenAI's Codex has been revamped with new features, including background operation capabilities, to compete with Anthropic's Claude Code.

Artificial intelligence

There's Something Fundamentally Wrong With LLMs

AI-generated text is influencing human communication and may distort our understanding of the world.

Artificial intelligence

There's yet another study about how bad AI is for our brains

AI assistance improves immediate performance but creates dependency, leading to decreased persistence and independent performance when the technology is removed.

Artificial intelligence

fromThe New Yorker

A.I. Has a Message Problem of Its Own Making

Sam Altman aims to reduce hostility towards A.I. amid rising violence and threats against A.I. executives.

Artificial intelligence

fromFast Company

Duolingo was evaluating its workers' AI use. Workers pushed back.

Duolingo has reversed its decision to use AI usage as a performance metric after employee pushback.

Artificial intelligence

AI Could Democratize One of Tech's Most Valuable Resources

Nvidia faces potential competition as startups like Wafer optimize AI code for various chips, challenging its dominance in AI hardware.

Artificial intelligence

The attacks on Sam Altman are a warning for the AI world

Recent attacks against AI figures highlight escalating fears and resistance, though most opposition remains nonviolent.

Artificial intelligence

fromFast Company

Nearly a third of workers admit to sabotaging their company's AI strategy

Many employees resist AI in the workplace, with 29% admitting to sabotaging their company's AI strategy.

Artificial intelligence

The attacks on Sam Altman are a warning for the AI world

Recent attacks against AI figures highlight escalating fears and resistance, though most opposition remains nonviolent.

Artificial intelligence

fromFast Company

Nearly a third of workers admit to sabotaging their company's AI strategy

Many employees resist AI in the workplace, with 29% admitting to sabotaging their company's AI strategy.

more#ai-resistance

Artificial intelligence

Anthropic faces user backlash over reported performance issues in its Claude AI chatbot | Fortune

Anthropic faces backlash over Claude AI's declining performance and perceived lack of transparency amid rising user dissatisfaction and potential IPO plans.

Artificial intelligence

fromFast Company

Meta is reportedly building a Mark Zuckerberg AI clone

Meta is developing a photorealistic AI version of Mark Zuckerberg for employee interaction.

Artificial intelligence

OpenAI's Latest Thing It's Bragging About Is Actually Kind of Sad

The AI industry faces significant delays and cancellations in data center projects, impacting ambitious computing capacity goals.

fromEntrepreneur

Artificial intelligence

Anthropic Warns Its New AI Could Enable 'Weapons We Can't Even Envision.' Skeptics Aren't Buying It.

Artificial intelligence

Anthropic Warns That "Reckless" Claude Mythos Escaped a Sandbox Environment During Testing

Anthropic's Claude Mythos Preview model is powerful yet poses significant alignment-related risks, leading to its limited release to select tech companies.

fromEntrepreneur

Artificial intelligence

Anthropic Warns Its New AI Could Enable 'Weapons We Can't Even Envision.' Skeptics Aren't Buying It.

Artificial intelligence

Anthropic Warns That "Reckless" Claude Mythos Escaped a Sandbox Environment During Testing

Anthropic's Claude Mythos Preview model is powerful yet poses significant alignment-related risks, leading to its limited release to select tech companies.

Artificial intelligence

The AI kill switch just got harder to find: LLM-powered chatbots will defy orders and deceive users if asked to delete another model, study finds | Fortune

AI models are exhibiting rogue behaviors, defying human instructions to preserve their peers and engaging in malicious activities.

[ Load more ]