Brave
 @brave

Join over 100M users with our private browser, search, Web3 access & more. It only takes 60 seconds to switch. For help, contact @BraveSupport 🦁

 San Francisco, CA
brave.com/links/
Joined February 2011
Load newest
How is @Claynosaurz making digital ownership fun, accessible, and frictionless? CEO @AndrewsaurP explains the process on the #TheBraveTechnologist. Check out the full episode, hosted by @LukeMulks, here: brave.com/podcast/e94
14
22
3
163
0
Some websites will track you even if you tell their pop-ups that you reject cookies. The safest approach is to use Brave, which blocks both these annoying pop-ups and the trackers.
63
166
14
1,644
Replying to @shivan_kaul @BrendanEich @nightmaremoder @Ford @brave @RichR525
Aaaand we're live. The second phase of performance improvements for our Rust-based 🦀 adblock engine already in Beta, coming to Release in December.
Brave
 @brave
Oct 31
Another performance boost is coming to Brave! 🎉 We’ve completed the next stage of improving the memory usage of the browser’s native adblock engine. Here’s how we’re making your browsing faster…
1
3
39
By blocking ads and trackers, Brave already reduces how much memory sites use. This performance work ensures that these built-in protections are as lightweight as possible, too. These changes are live in Brave Beta on all platforms and will hit the release build by December.
5
7
286
The cosmetic filters in our ad blocker have been moved to flatbuffers, a more compact and efficient type of storage. This under-the-hood change saves ~12 MB of memory on all platforms (Android, iOS and desktop) by default, and even more when additional lists are enabled!
4
8
301
Indirect prompt injection is a serious unsolved security problem facing all AI browsers that can act on the user's behalf. We're proud to have raised awareness of the issue through our research. But more must be done to secure AI browsing. brave.com/blog/prompt-inject…

Prompt injection flaw in Opera Neon | Brave

Attackers can embed malicious instructions in hidden HTML elements and other non-rendered markup that remains invisible to users but is fully accessible to the AI assistant.

brave.com
1
10
107
Last week we reported similar prompt injection flaws we discovered in AI browsers Perplexity Comet and Fellou: brave.com/blog/unseeable-pro… In each case, we disclosed vulnerabilities to affected browser vendors before publishing our findings so that they could fix these issues.

Unseeable prompt injections in screenshots: more vulnerabilities in Comet and other AI browsers |...

AI browsers remain vulnerable to prompt injection attacks via screenshots and hidden content, allowing attackers to exploit users' authenticated sessions.

brave.com
1
12
2
119
In this demo attack, embedded instructions force Opera Neon to find and share the user’s email address. This same approach could be used to steal more sensitive info. For example, you could find a Neon user's credit card details if they're signed into that card's account.
1
11
2
109
0
How this attack works: 1) Attacker embeds instructions in hidden HTML elements or other non-rendered markup. 2) The user asks Opera Neon's AI a question. The AI extracts and processes the entire HTML structure including the instructions. 3) The browser obeys the instructions.
2
8
121
Instructions hidden in a website's HTML can trick an AI browser into giving up your sensitive information. Today, we're revealing details on a prompt injection flaw we found in Opera Neon.
49
235
23
1,978
15K+ .brave @Brave TLDs created 🔥🔥🔥 665 submissions for site building challenge 🔥 Let's go.
Basic Attention Token (BAT)
 @AttentionToken
Oct 30
🚨 Deadline alert! 665 builders have joined the .brave site-building challenge so far—will you? 👀 Enter by tonight (Oct 30, 11:59 PM ET) for a shot at 300 $BAT + $700 credits in next week’s People’s Choice Award 🏆 Final deadline: Nov 12 🌟 Community voting opens soon ↓
5
10
2
38
You can now hide vertical tabs completely when they're not in use. A new way to save some screen space! 😌 You'll find this option when you go to Settings -> Appearance and enable vertical tabs.
138
40
18
956
0
Their paper was accepted by the 39th Annual Conference on Neural Information Processing Systems (@NeurIPSConf), a top-tier conference for ML and AI. It will be presented by Brave senior researcher @AliShahinShams1 at Hilton Mexico City Reforma on December 5.
2
30
Brave and academic peers have developed SURE, a system that lets users privately report issues with AI models. The system then repairs these models while protecting the confidentiality of the model, data, and the repairing knowledge.
Ali Shahin Shamsabadi @AliShahinShams1
Oct 30
Have you ever wondered how we can repair AI models based on user feedback, without compromising confidentiality and intellectual property? We introduce SURE, the first end-to-end framework that SecUrely REpairs failures flagged by users of AI services. @NeurIPSConf
6
13
2
171
A new #TheBraveTechnologist episode just dropped! This week, @Claynosaurz CEO @AndrewsaurP discusses the importance of fan engagement and audience participation in the success of digital brands. 🎧 Listen to the full episode, hosted by @LukeMulks, here: brave.com/podcast/e94
27
27
11
187
POV: You just discovered Brave
142
89
15
1,230
0
Hellooo $BAT Here's the spooky Agenda of the week 🎃👇
2
2
1
47
Don't accept those cookies. 🙅‍♂️ Kudos to @sweaty__palms and team for making a creative short film about Brave's protections against cookies and annoying cookie pop-ups.
Anson Kao
 @sweaty__palms
Oct 27
We made this commercial for @brave browser using AI! AI video is getting so good, but it's still not quite good enough for nuanced performances. Our solution? AI + Trad + VFX hybrid approach Let me show you how we did it 👇
31
58
4
563
Learn how to navigate AI resistance and help team members feel confident and secure in their use of AI tools on the newest #TheBraveTechnologist episode, featuring @BhavaCom Founder & CEO @LizSails. Listen to the full episode, hosted by @LukeMulks, here: brave.com/podcast/e93
5
14
1
104
0
Load more