Pinned Tweet
Seriously - the only attraction for me to do this, is the ability for me to be able to say it in my words, when I want, and how I want. Yes, I will make sure to the best of my ability to follow rules re: public company fiduciaries and dissemination of information. But the idea that someone else crafts your communication and tweets for you - is not better than giving an interview to a publication and worry that the journalist will cherry pick to validate their narrative. Control the narrative and the consequences by doing it yourself.
@nikesharora @HarryStebbings You're one of the few large company CEOs that seems to be writing their own tweets and they are interesting so you should continue
55
11
528
145,512
Lol.
We all want our LLMs to respond faster, to be smarter, to not hallucinate. Agents have just made an appearance on the scene, wait till every app is rewritten on your iPhone and it's interacting with agents. Enterprises are in the first innings of AI adoption, coding is diffusing and promises to continue to grow. As models get smarter, it will open up more applications and use cases - so yes demand is insatiable and it has nothing to do with security. Of course, security will introduce trust to enterprises who want to adopt AI.
Have you tried to get your hands on compute? CPU and GPU? It takes 3/4 times the spend to deliver a MW compared to 2 years ago. There send to be no respite for another 24 months at least, hoping compression techniques, advances on technology will relieve some of the pressure and things normalize in due course.
We all want our LLMs to respond faster, to be smarter, to not hallucinate. Agents have just made an appearance on the scene, wait till every app is rewritten on your iPhone and it's interacting with agents. Enterprises are in the first innings of AI adoption, coding is diffusing and promises to continue to grow. As models get smarter, it will open up more applications and use cases - so yes demand is insatiable and it has nothing to do with security. Of course, security will introduce trust to enterprises who want to adopt AI.
Have you tried to get your hands on compute? CPU and GPU? It takes 3/4 times the spend to deliver a MW compared to 2 years ago. There send to be no respite for another 24 months at least, hoping compression techniques, advances on technology will relieve some of the pressure and things normalize in due course.
@firesidealphaa@nikesharoraorks generates massive revenue by securing AWS, Azure, and Google Cloud environments. By publicly positioning these hyperscalers as the ultimate power brokers of the AI era, the "owners of scarce compute", he aligns perfectly with his most critical go-to-market partners.
If compute is an incredibly scarce, highly contested resource, protecting the cloud architectures housing it becomes an existential priority for enterprises. This narrative justifies massive security budgets and premium pricing for cloud posture management platforms to lock down high-value AI workloads.
Describing demand as "insatiable" and compute as "scarce" accelerates enterprise anxiety. Companies rushing to secure their cloud footprint and deploy models before they are priced out of the GPU rental market will rapidly expand their cloud infrastructure, directly pulling security spending forward with it.
26
21
244
34,924
@nikesharora I understand this would be out of your support terms and so on, but would you mind sending an advisory out pointing to this vulnerability and a fix if possible with high priority emails out to all your customers?
People tend to take Palo Alto security advisories seriously.
Thank you. @nikesharora
8
9
142
29,803
You know it's serious when heads of state get involved.
皆様に、今夜に対応して頂きたい大至急のお願いです。
多くのウェブサイトで使用されているPHPに関し、PHPのセキュリティチームが、重大な２件の脆弱性に関する注意喚起を行っています。
公開しているウェブサイトでPHPを使用しているか確認し、今夜、10月21日午後11時（日本時間）に公開される修正パッチについて、可及的速やかに適用できるよう準備をお願いします。
修正パッチを速やかに適用することが困難な場合は、例えばウェブサイト等を一時的に停止させることも、リスクを低減させるための有効策の一つと考えられます。
国家サイバー統括室のXアカウントでも関係する情報を発信していますので、そちらも参考にして下さい。
国家サイバー統括室(注意・警戒情報)（@cyber_forecast） / X
18
77
991
118,694
Will have to try harder or @michaelpwalrath and I will settle it on the golf course :).
@bdistel @nikesharora Damn. Thanks….I think. 😜
2
1
20
15,148
You should read my sequel to the ninja move post. I suggested that the original ninja move could actually backfire if we didn't change our approach as an industry to focus on benefits and instead turn to delivering this technology safely, securely yet still at pace. It is not possible to pace open source even if you got some frontier labs to collaborate switch would continue to define the pace.
Today, Nikesh decided not to directly address the question @davidfaber from @CNBC posed about Nikesh's X posts. This is the move of an experienced executive, avoiding a PR nightmare by staying out of muddy ground. It also allowed Nikesh to stay on message for the interview, the message that he and Jay Clayton wanted to get across.
Not sure it did much to convince the American people that Nikesh's view has changed since these posts, and who would expect them to after only a few weeks. Of course, most of the people watching #CNBC probably care more about the AI trade staying in place, than some abstract concept of safety that few can probably understand anyway. One of Nikesh's previous criticisms of people who had raised AI safety concerns was that it might derail the #AI trade. Given security companies were impacted by the #Saasapocalypse, that might be understandable.
Nonetheless, it is hard to escape the perception that the people leading the #superintellgence effort, continue to have mixed feelings, at best, about the people who got this whole conversation started.
#security #safety #alignment
youtu.be/J-zZq7x9Jj0?si…
6
4
31
11,817
Free wisdom for those of you opining on my last tweet. It's easy to cast individuals away because you disagree with something they did or their opinions. When I turned 40, I turned a corner and decided to appreciate the good in someone for their act, for the moment and for what they did. My friends are my friends and I respect many people for their achievements in various fields, but I wouldn't have them work on stuff In other areas. An absolutist bar on each human is a recipe for failure. I expect things from people, but I don't expect them to be perfect. If their heart is in the right place, and their intent on the matter at hand is well placed, I respect the effort.
36
24
608
58,160
Cybersecurity’s AI reckoning has arrived, and @PaloAltoNtwks CEO @nikesharora has become a go-to adviser to CEOs. bit.ly/4z16ena
34
40
611
1,634,749
The meeting at the WH on SI was an epic moment. The good and the great of AI (SI) showed up from all over, I was just happy to be in the room.
First and foremost I felt that @realDonaldTrump showed us what a true leader is. He was thoughtful and focussed on ensuring that SI was developed at pace and was going to be deployed safely and securely.
He encouraged the room to focus on the positive impacts of technology, demonstrate the true benefits of the technology for each and every one out there.
His message was clear - engage with communities get grassroots support for what is going to be a pivotal moment for America and the world.
He was statesmanlike and clearly focussed on the US winning the SI race.
The focus needs to be how to ensure what is being built is safe and useful. There are enough laws and regulatory frameworks that allow for effective governance we don't need more. What we need is:
1. Accountability by the labs
2. we need SI to have the right APIs available for cybercontrols and third party governance and
3. We need equal focus on diffusion of this technology in a useful and effective manner.
We all walked away with continued and renewed enthusiasm to bring on the SI future.
When you get cyber companies focussing on furthering a technology you have already moved the needle.
First and foremost I felt that @realDonaldTrump showed us what a true leader is. He was thoughtful and focussed on ensuring that SI was developed at pace and was going to be deployed safely and securely.
He encouraged the room to focus on the positive impacts of technology, demonstrate the true benefits of the technology for each and every one out there.
His message was clear - engage with communities get grassroots support for what is going to be a pivotal moment for America and the world.
He was statesmanlike and clearly focussed on the US winning the SI race.
The focus needs to be how to ensure what is being built is safe and useful. There are enough laws and regulatory frameworks that allow for effective governance we don't need more. What we need is:
1. Accountability by the labs
2. we need SI to have the right APIs available for cybercontrols and third party governance and
3. We need equal focus on diffusion of this technology in a useful and effective manner.
We all walked away with continued and renewed enthusiasm to bring on the SI future.
When you get cyber companies focussing on furthering a technology you have already moved the needle.
66
107
958
59,611
lnkd.in/p/g-4GA-TY ok that's funny :). Almost Friday.
13
1
84
17,798
Conceptually you can access APIs for control panels in the enterprise use case. The API and MCP is less of an issue - the access, audit, ownership of actions and liability is. In the enterprise use case, one who accesses is responsible.
@nikesharora Couldn't agree more.
@nikesharora do you take the same approach for Palo products? E.g will an "enterprise muse" be able to configure NGFW?
20
4
70
21,596
This aged well, Muse and Instinct took longer than expected.
Lol. A string of connected tasks. Like book me a trip and a restaurant for lunch in one task. So you have to interface with two back ends. My point is, we will need an agent exchange, broker, or rules of engagement. Our current app owners will fight for owning the customer and resist agent APIs.
16
6
303
76,710
This will be a bigger battle than anyone anticipates. It is only a matter of time before there is an Apple and Google version of Muse and possibly TikTok, in addition to the frontier LLM agents. Maybe a commerce agent from Amazon.
Every app that is a services, marketplace or commerce app will need to existentially decide to open APIs for consumer agents to interact. Smaller players have no choice. Ad revenues are more than transaction fees, either the consumer benefits or distribution aggregators will demand a higher transaction fare.
I know I don't want an agent for each app. I would like my agent to be able to do tasks I require. We can already see consumers getting trained on that behavior by the frontier labs.
Those with network moats - restaurants, groceries, drivers might be able to withstand for a while, over time convenience and end user experience will win and they will have to align. Content moats (protected by copyright) could decide to allow agents or chose to hold on to the consumer interaction. I suspect other than the feeling of a lack of control, it won't change their economics.
Commoditized back ends will need to worry, insurance, tickets, hotels, services - if they don't adapt new players will.
Every app that is a services, marketplace or commerce app will need to existentially decide to open APIs for consumer agents to interact. Smaller players have no choice. Ad revenues are more than transaction fees, either the consumer benefits or distribution aggregators will demand a higher transaction fare.
I know I don't want an agent for each app. I would like my agent to be able to do tasks I require. We can already see consumers getting trained on that behavior by the frontier labs.
Those with network moats - restaurants, groceries, drivers might be able to withstand for a while, over time convenience and end user experience will win and they will have to align. Content moats (protected by copyright) could decide to allow agents or chose to hold on to the consumer interaction. I suspect other than the feeling of a lack of control, it won't change their economics.
Commoditized back ends will need to worry, insurance, tickets, hotels, services - if they don't adapt new players will.
Amazon cuts off Muse.
While I am bullish Meta and Muse, I think many people are overlooking the digital knife fight that’s about to occur
Nobody wants to get commoditized or layered here. Let the games begin
178
255
2,262
3,913,622
Well said. Agree.
@nikesharora is exactly where alignment gets difficult. It’s not just about preventing the model from jumping guardrails. It’s about setting the right constraints so it consistently chooses to operate within them.
But then comes the harder question: whose rules?
Easy examples create false confidence because most of us agree on the answer. The real alignment problem lives in edge cases where values conflict, context matters, and reasonable people disagree.
That’s where I think alignment stops being purely a model problem and starts becoming a governance problem.
5
0
27
16,085
The recent flurry of math problem solved?
@nikesharora Name one bleeding edge research example
3
0
31
12,025
Lolol.
@nikesharora Let's merge OpenAI - Anthropic - xAI and make Nikesh Arora the CEO of it. The above is a CEO-Statesman-Yoda wisdom. @nikesharora
15
4
159
32,913
Alignment is less about jumping guardrails and setting the right constraints. It's about making sure the model wants to play by rules. The question becomes "whose rules". Think of edge cases. Your example is easy, we can all suggest the right answer here.
@nikesharoraikesh! I'd push on your point 5.
An enterprise shouldn’t have to wait for agreement on whose values a model should align to before deciding what an agent is allowed to do.
Take a coding agent for example: permission to investigate a production issue shouldn’t automatically mean permission to change access controls or delete data. Those boundaries need to be enforced outside the model, even when it gives a convincing explanation for crossing them. That doesn’t solve alignment, but it gives enterprises a concrete way to limit the consequences of mistakes as capabilities improve.
I’d also be careful about discouraging labs from sharing failures. We need that visibility. What I’d ask is that every disclosure include what failed, what changed, and how they tested the fix. It just can't be an RSI blackbox.
11
1
45
12,606
Small group may not be AI labs but complimentary players.
@nikesharora point 3 says a few leaders self-pacing is illusory because one defector restarts the race, but the fix you land on is also solve it with a few players and demonstrate leadership so others follow. what makes small-group coordination hold for alignment if it cannot hold for pacing?
1
1
18
11,011