Close Menu
TechTost
  • AI
  • Apps
  • Crypto
  • Fintech
  • Hardware
  • Media & Entertainment
  • Security
  • Startups
  • Transportation
  • Venture
  • Recommended Essentials
What's Hot

HomeBoost’s app will show you where you can save money on your utility bills

Tesla profits down 46% in 2025

Nvidia CEO refutes report that his company’s $100 billion OpenAI investment has stalled

Facebook X (Twitter) Instagram
  • About Us
  • Contact Us
  • Privacy Policy
  • Terms and Conditions
  • Disclaimer
Facebook X (Twitter) Instagram
TechTost
Subscribe Now
  • AI

    Nvidia CEO refutes report that his company’s $100 billion OpenAI investment has stalled

    31 January 2026

    A look at Physical Intelligence, the startup building Silicon Valley’s busiest robot brains

    31 January 2026

    Anthropic brings agent plug-ins to Cowork

    30 January 2026

    Guys, I don’t think Tim Cook knows how to monetize AI

    30 January 2026

    SpaceX, Tesla and Elon Musk’s xAI are in talks to merge, according to reports

    29 January 2026
  • Apps

    WhatsApp will now charge for AI chatbots to operate in Italy

    31 January 2026

    Bluesky issues its first transparency report, noting an increase in user reports and legal requirements

    31 January 2026

    Instagram may soon let you remove yourself from someone’s close friends list

    30 January 2026

    Google Maps now lets you access Gemini while walking and cycling

    30 January 2026

    OpenAI’s Sora app is struggling after its stellar launch

    29 January 2026
  • Crypto

    Hackers stole over $2.7 billion in crypto in 2025, data shows

    23 December 2025

    New report examines how David Sachs may benefit from Trump administration role

    1 December 2025

    Why Benchmark Made a Rare Crypto Bet on Trading App Fomo, with $17M Series A

    6 November 2025

    Solana co-founder Anatoly Yakovenko is a big fan of agentic coding

    30 October 2025

    MoviePass opens Mogul fantasy league game to the public

    29 October 2025
  • Fintech

    How Sequoia-backed Ethos went public while rivals lagged behind

    30 January 2026

    5 days left for TechCrunch Disrupt 2026 +1 pass with 50%

    26 January 2026

    50% off +1 ends | TechCrunch

    23 January 2026

    Capital One acquires Brex for a steep discount to its valuation, but early believers are laughing all the way to the bank

    23 January 2026

    Tiger Global and Microsoft will fully exit Walmart-backed PhonePe through its IPO

    22 January 2026
  • Hardware

    Microsoft won’t stop buying AI chips from Nvidia, AMD even after its own is released, says Nadella

    30 January 2026

    The iPhone just had its best quarter ever

    30 January 2026

    Snap is serious about specs, spinning off AR glasses into a standalone company

    28 January 2026

    Android phones are getting more anti-theft features

    27 January 2026

    Apple’s new AirTag is stronger and can be found at greater distances

    26 January 2026
  • Media & Entertainment

    OnlyFans is considering selling a majority stake to Architect Capital

    31 January 2026

    Last 24 hours to get 50% off +1 pass for Disrupt 2026 | TechCrunch

    30 January 2026

    Disrupt 2026: +1 cards are almost gone with only 3 days left

    28 January 2026

    Sci-fi writers, Comic-Con say goodbye to artificial intelligence

    26 January 2026

    Amagi debuts in India as cloud TV software company tests investor appetite

    24 January 2026
  • Security

    Russian hackers breached Poland’s power grid thanks to poor security, report says

    31 January 2026

    Whistleblower Told FBI Jeffrey Epstein Had ‘Personal Hacker’

    31 January 2026

    Fintech firm Marquis blames hack on firewall provider SonicWall for data breach

    30 January 2026

    Apple’s new iPhone and iPad security feature restricts mobile networks from collecting accurate location data

    29 January 2026

    If you live in the UK, you will probably no longer be able to visit Pornhub

    29 January 2026
  • Startups

    HomeBoost’s app will show you where you can save money on your utility bills

    1 February 2026

    Qualcomm backs SpotDraft to scale AI with on-device deal doubling valuation to $400 million

    31 January 2026

    Redwood Lands Google for $425M Series E as AI Power Needs Grow

    31 January 2026

    Tiny startup Arcee AI built a 400B parameter open source LLM from scratch to best Meta’s Llama

    30 January 2026

    Upwind Raises $250M in $1.5B Valuation to Continue Building ‘Runtime’ Cloud Security

    30 January 2026
  • Transportation

    Tesla profits down 46% in 2025

    1 February 2026

    Waymo robotaxi hits a child near an elementary school in Santa Monica

    31 January 2026

    Tesla’s energy storage business is growing faster than any other part of the company

    30 January 2026

    Waymo robotaxis now offers rides to and from San Francisco International Airport

    30 January 2026

    Tesla to invest $2 billion in Elon Musk’s xAI

    29 January 2026
  • Venture

    a16z contributor Kofi Ampadu will be leaving permanently after the TxO program is discontinued

    31 January 2026

    Reid Hoffman urges Silicon Valley leaders to stop bending the knee to President Trump

    31 January 2026

    VC 2150 raises €210 million to solve cities’ climate challenges

    27 January 2026

    Obvious Ventures lands fund five with a 360-degree view of planetary, human and financial health

    27 January 2026

    Vinod Khosla publicly disavows Keith Rabois’ comments on ICE shooting

    26 January 2026
  • Recommended Essentials
TechTost
You are at:Home»Security»OpenAI says AI browsers can always be vulnerable to injection attacks
Security

OpenAI says AI browsers can always be vulnerable to injection attacks

techtost.comBy techtost.com23 December 202505 Mins Read
Share Facebook Twitter Pinterest LinkedIn Tumblr Email
Openai Says Ai Browsers Can Always Be Vulnerable To Injection
Share
Facebook Twitter LinkedIn Pinterest Email

Although OpenAI is working to harden its Atlas AI browser against cyberattacks, the company admits that just-in-time injections, a type of attack that manipulates AI agents into following malicious instructions often hidden in web pages or emails, is a risk that won’t go away anytime soon — raising questions about how secure AI agents can operate on the open web.

“Direct injection, like web scams and social engineering, is unlikely to be fully ‘solved’,” OpenAI wrote on Monday. blog post detailing how the company is beefing up Atlas’ armor to combat the relentless attacks. The company admitted that the “agent mode” in ChatGPT Atlas “expands the security threat surface”.

OpenAI released its ChatGPT Atlas browser in October, and security researchers were quick to publish their demos, showing that it was possible to type a few words into Google Docs that could change the behavior of the underlying browser. Same day, Brave published a blog post explaining that indirect direct injection is a systematic challenge for AI-powered browsers, including Perplexity’s Comet.

OpenAI isn’t alone in recognizing that timely injections aren’t going away. THE The UK’s National Cyber ​​Security Center warned earlier this month that direct injection attacks against genetic AI applications “can never be fully mitigated,” putting websites at risk of falling victim to data breaches. The UK government agency has advised cyber professionals to reduce the risk and impact of timely injections, rather than believing that attacks can be “stopped”.

For OpenAI’s part, the company said: “We see direct injection as a long-term AI security challenge, and we should continually strengthen our defenses against it.”

The company’s response to this Sisyphean task? A proactive rapid response cycle that the company says shows early promise for helping discover innovative attack strategies internally before they are exploited “in the wild.”

That’s not entirely different from what competitors like Anthropic and Google have been saying: that to combat the persistent threat of just-in-time attacks, defenses must be layered and constantly stress-tested. Recent work by Googlefor example, it focuses on architectural and policy-level controls for agent systems.

But where OpenAI takes a different tactic is with the “LLM-based automated attacker”. This attacker is basically a bot that OpenAI trained, using reinforcement learning, to play the role of a hacker looking for ways to sneak malicious instructions into an AI agent.

The bot can test the attack in a simulation before actually using it, and the simulator shows how the AI ​​target would think and what actions it would take if it saw the attack. The bot can then study this response, modify the attack and try again and again. This picture of the target AI’s internal reasoning is something outsiders don’t have access to, so in theory OpenAI’s bot should be able to find flaws faster than a real-world attacker.

It’s a common tactic in AI security testing: create an agent to find edge cases and quickly test them in a simulation.

“Us [reinforcement learning]”A trained attacker can direct an agent to execute sophisticated, long-horizon malicious workflows that unfold over tens (or even hundreds) of steps,” OpenAI wrote. “We also observed new attack strategies that did not appear in the human clustering campaign or in external reports.”

Image Credits:OpenAI

In a demonstration (pictured in part above), OpenAI showed how its automated attacker cracked a malicious email in a user’s inbox. When the AI ​​agent later scanned the inbox, it followed the hidden instructions in the email and sent a resignation message instead of composing an out-of-office reply. However, after the security update, “agent mode” was able to successfully detect the direct injection attempt and flag it to the user, according to the company.

The company says that while timely injection is difficult to ensure infallibly, it relies on large-scale testing and faster patch cycles to harden its systems before they are exposed to real-world attacks.

An OpenAI spokesperson declined to share whether the Atlas security update has resulted in a measurable reduction in successful injections, but says the company has been working with third parties to harden Atlas against direct injection since before the release.

Rami McCarthy, principal security researcher at cybersecurity firm Wiz, says reinforcement learning is a way to continuously adapt to attacker behavior, but it’s only part of the picture.

“A useful way to think about risk in AI systems is autonomy multiplied by access,” McCarthy told TechCrunch.

“Browser agents tend to sit in a difficult part of this space: moderate autonomy combined with very high access,” McCarthy said. “Many current recommendations reflect this trade-off. Limiting connected access primarily reduces exposure, while requiring verification of confirmation requests limits autonomy.”

Those are two of OpenAI’s recommendations to users to reduce their own risk, and a spokesperson said Atlas is also trained to get user confirmation before sending messages or making payments. OpenAI also suggests that users give agents specific instructions, rather than giving them access to your inbox and telling them to “do whatever it takes.”

“High latitude makes it easier for hidden or malicious content to affect the agent, even when safeguards are in place,” according to OpenAI.

While OpenAI says protecting Atlas users from timely injections is a top priority, McCarthy raises some skepticism about the return on investment for vulnerable browsers.

“For most everyday use cases, agent browsers don’t yet offer enough value to justify their current risk profile,” McCarthy told TechCrunch. “The risk is high given their access to sensitive data like email and payment information, even though that access is also what makes them powerful. That balance will evolve, but today the trade-offs are still very real.”

AI browser atlas atlas chatgpt attacks Browsers cyber security injection OpenAI rapid injections vulnerable
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
Previous ArticleSplat’s app uses artificial intelligence to turn your photos into coloring pages for kids
Next Article Paramount renews bid for Warner Bros, securing $40 billion backing from Larry Ellison
bhanuprakash.cg
techtost.com
  • Website

Related Posts

Nvidia CEO refutes report that his company’s $100 billion OpenAI investment has stalled

31 January 2026

Russian hackers breached Poland’s power grid thanks to poor security, report says

31 January 2026

Whistleblower Told FBI Jeffrey Epstein Had ‘Personal Hacker’

31 January 2026
Add A Comment

Leave A Reply Cancel Reply

Don't Miss

HomeBoost’s app will show you where you can save money on your utility bills

1 February 2026

Tesla profits down 46% in 2025

1 February 2026

Nvidia CEO refutes report that his company’s $100 billion OpenAI investment has stalled

31 January 2026
Stay In Touch
  • Facebook
  • YouTube
  • TikTok
  • WhatsApp
  • Twitter
  • Instagram
Fintech

How Sequoia-backed Ethos went public while rivals lagged behind

30 January 2026

5 days left for TechCrunch Disrupt 2026 +1 pass with 50%

26 January 2026

50% off +1 ends | TechCrunch

23 January 2026
Startups

HomeBoost’s app will show you where you can save money on your utility bills

Qualcomm backs SpotDraft to scale AI with on-device deal doubling valuation to $400 million

Redwood Lands Google for $425M Series E as AI Power Needs Grow

© 2026 TechTost. All Rights Reserved
  • About Us
  • Contact Us
  • Privacy Policy
  • Terms and Conditions
  • Disclaimer

Type above and press Enter to search. Press Esc to cancel.