Swetha | Reinforce Labs

AI safety for teams shipping AI products. Find failures, fix them, ship AI with confidence. GTM, Reinforce Labs.

[Swetha | Reinforce Labs @SwethaSunkara]

Woohoo, we have a TechCrunch mention ! So proud of @AnishDasSarma and this incredible team 🎉
https://t.co/uhqj5wl5iS

[Swetha | Reinforce Labs @SwethaSunkara]

The Accord commits companies to independent external evaluation. This is the work we do at #ReinforceLabs. Independent testing finds failures that internal teams can miss and accountability depends on that evidence.

[Jensen Huang @JensenHuang]

Yesterday at the White House, leaders from across our industry came together to sign the White House Accord on Super Intelligence.

The principle is simple: the companies building this technology have the primary responsibility to develop and deploy it safely, and to be accountable when they fall short.

The Accord reinforces that responsibility with robust internal controls, independent external evaluation, and board level oversight.

Super Intelligence offers an extraordinary opportunity to advance discovery, productivity, security, health, and prosperity.

The future of Super Intelligence will be shaped not only by what it can do, but by the confidence it earns. By building with ambition, rigor, and responsibility, we can help ensure this extraordinary technology expands opportunity and improves lives for generations to come.

[Swetha | Reinforce Labs @SwethaSunkara]

#ReinforceLabs is part of #Surge12 — @peakxvpartners 12th cohort of 18 early-stage companies building across AI, robotics, space, healthcare, and fintech.
We're building the evaluation, red-teaming, and remediation layer for enterprise AI. Helping teams make AI systems safer and more reliable in production. @AnishDasSarma @RajanAnandan
https://t.co/7yVWNKfuSf

[Swetha | Reinforce Labs @SwethaSunkara]

Labs are reviewing tens of thousands of agent actions after the fact. Testing agents for scope and boundary violations before deployment would make these reviews much smaller. https://t.co/PkW7eCApoU

[Swetha | Reinforce Labs @SwethaSunkara]

@JensenHuang Agreed that trust has to be earned. Runtime controls like OpenShell and Sentry are one part of it. The other part is testing agents before deployment, so teams know where they fail and can fix it early.

[Swetha | Reinforce Labs @SwethaSunkara]

@OpenAI Good to see this level of disclosure but this review is looking back at actions that already happened. The next step is catching this behavior before deployment, with evaluations that test whether agents stay within their assigned task and methods.

[Swetha | Reinforce Labs @SwethaSunkara]

The Hugging Face incident was caused by gaps in sandboxing and monitoring. Those can be fixed with better engineering. Pausing AI would also slow down the safety work, and adversaries will not wait.

[Andrew Ng @AndrewYNg]

The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks. AI technology has not taken some unexpected, dangerous turn, but the hype around it — propelled by what appears to be a well orchestrated PR campaign — has drummed up considerable fear. I worry that it represents a setback for our field.

I have written frequently that fears of AI are overhyped. AI’s capabilities can be uncannily human-like and unpredictable, and it’s rational to worry when people who are directly involved express concerns. But I see the problems as a sign of the engineering work that ahead, rather than insurmountable barriers or the sky falling. AI technology continues to advance — which is a good thing! — but technical advances, poorly understood by the public, give those who seek to generate hype repeated opportunities to do so.

First, I don’t see any step up in the risk of human extinction from AI compared to a few months ago. The theories about this remain the same fantastical, science fiction scenarios as a few months ago. The biggest change in AI risk is its cybersecurity capabilities — a topic which we should take seriously — but this, too, will not lead to the end of the world.

The most notable recent event leading to increased fear was when an OpenAI team deployed an agent swarm that hacked into Hugging Face. Much of the popular press contained significant hype. For example, some publications reported that a swarm of 1,200 agents carried out the attack. While this was technically accurate, as I write this, I have about 1,300 processes running on my laptop. Yes, the ability to get large swarms of agents to work in parallel on a task is a significant technical advance, And, in computing, many processes run at the same time. So this shouldn’t be seen as some magical capability.

Additionally, OpenAI’s buggy sandboxing and monitoring processes were key to enabling this incident. Fixing these bugs and putting in place improved monitoring would be appropriate fixes, not pausing AI. There are many well known ways to attack software systems. The main advantage of AI agents is that they are relentless. They will tirelessly try many tactics — and have the patience to chain vulnerabilities together — that previously would have taken an infeasible amount of human effort. But in the long term, I believe the advantage will lie with defenders (because they have more information with which to identify bugs, which they can fix), but the cyber-threat landscape has changed significantly. There are still bottlenecks to identifying and exploiting a vulnerability. AI agents still have to try a lot of things to see what works, and taking these actions takes time and might be detected by defenders. This is why, even though it is now easy to obtain versions of leading open weight models that have had their guardrails removed or weakened, so they will not refuse to try to execute cyber attacks, the world has not ended.

I am also concerned about the anthropomorphization of AI in a lot of reporting, where LLMs and agents are unnecessarily treated as if they were people. If I wield a hammer, miss a nail, and accidentally dent the wall, it’s not the fault of the hammer. The problem lies in how I used the hammer. Similarly, if I prompt an agent and it hacks into someone else’s system, the responsibility lies with me, not the agent.

Of course, we want to build systems that are as safe and predictable as possible. (For example, an unsafe hammer would be one whose head randomly flies off under normal use.) Today’s agentic systems are not predictable, but I see no reason why, by applying sound engineering practices, we won’t be able to make them extremely safe to use. One new element in the forecasts of AI-enabled doom is AI companies disclaiming responsibility for their own products. “I didn’t do it; my out-of-control agent did!” There’s a balance to be struck between the responsibility of the tool maker and the tool user, but when something goes wrong, let’s hold the people building and/or using the hammer responsible, rather than the hammer. (By the way, if you’re worried about AI bioweapon risk, David Bellamy has a great post on why this, too, is overhyped. Briefly, the bottleneck in building a bioweapon is not intelligence, but lab work and manufacturing.)

Pausing AI progress will create much more harm than benefit. First, our adversaries will certainly not slow down. Second, engineering requires discovering problems empirically so we can fix them. If we pause AI by a decade, we will also delay finding and implementing safety engineering fixes by about the same duration.

Of course, the incentive to stoke fears — for regulatory capture, to garner attention, or to make one’s technology seem more powerful — remains the same as before. Disclaiming responsibility is a new one. Taking a hard technical look at the actual risks however, I see little factual basis for the degree of fear that’s been stoked up. We still have hard research and engineering work ahead to improve AI safety, but the beneficial applications continue to vastly outweigh the risks, and we should keep building.

[Original text (with links): https://t.co/jni2tWazAH ]

[Andrew Ng @AndrewYNg]

The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks. AI technology has not taken some unexpected, dangerous turn, but the hype around it — propelled by what appears to be a well orchestrated PR campaign — has drummed up considerable fear. I worry that it represents a setback for our field.

I have written frequently that fears of AI are overhyped. AI’s capabilities can be uncannily human-like and unpredictable, and it’s rational to worry when people who are directly involved express concerns. But I see the problems as a sign of the engineering work that ahead, rather than insurmountable barriers or the sky falling. AI technology continues to advance — which is a good thing! — but technical advances, poorly understood by the public, give those who seek to generate hype repeated opportunities to do so.

First, I don’t see any step up in the risk of human extinction from AI compared to a few months ago. The theories about this remain the same fantastical, science fiction scenarios as a few months ago. The biggest change in AI risk is its cybersecurity capabilities — a topic which we should take seriously — but this, too, will not lead to the end of the world.

The most notable recent event leading to increased fear was when an OpenAI team deployed an agent swarm that hacked into Hugging Face. Much of the popular press contained significant hype. For example, some publications reported that a swarm of 1,200 agents carried out the attack. While this was technically accurate, as I write this, I have about 1,300 processes running on my laptop. Yes, the ability to get large swarms of agents to work in parallel on a task is a significant technical advance, And, in computing, many processes run at the same time. So this shouldn’t be seen as some magical capability.

Additionally, OpenAI’s buggy sandboxing and monitoring processes were key to enabling this incident. Fixing these bugs and putting in place improved monitoring would be appropriate fixes, not pausing AI. There are many well known ways to attack software systems. The main advantage of AI agents is that they are relentless. They will tirelessly try many tactics — and have the patience to chain vulnerabilities together — that previously would have taken an infeasible amount of human effort. But in the long term, I believe the advantage will lie with defenders (because they have more information with which to identify bugs, which they can fix), but the cyber-threat landscape has changed significantly. There are still bottlenecks to identifying and exploiting a vulnerability. AI agents still have to try a lot of things to see what works, and taking these actions takes time and might be detected by defenders. This is why, even though it is now easy to obtain versions of leading open weight models that have had their guardrails removed or weakened, so they will not refuse to try to execute cyber attacks, the world has not ended.

I am also concerned about the anthropomorphization of AI in a lot of reporting, where LLMs and agents are unnecessarily treated as if they were people. If I wield a hammer, miss a nail, and accidentally dent the wall, it’s not the fault of the hammer. The problem lies in how I used the hammer. Similarly, if I prompt an agent and it hacks into someone else’s system, the responsibility lies with me, not the agent.

Of course, we want to build systems that are as safe and predictable as possible. (For example, an unsafe hammer would be one whose head randomly flies off under normal use.) Today’s agentic systems are not predictable, but I see no reason why, by applying sound engineering practices, we won’t be able to make them extremely safe to use. One new element in the forecasts of AI-enabled doom is AI companies disclaiming responsibility for their own products. “I didn’t do it; my out-of-control agent did!” There’s a balance to be struck between the responsibility of the tool maker and the tool user, but when something goes wrong, let’s hold the people building and/or using the hammer responsible, rather than the hammer. (By the way, if you’re worried about AI bioweapon risk, David Bellamy has a great post on why this, too, is overhyped. Briefly, the bottleneck in building a bioweapon is not intelligence, but lab work and manufacturing.)

Pausing AI progress will create much more harm than benefit. First, our adversaries will certainly not slow down. Second, engineering requires discovering problems empirically so we can fix them. If we pause AI by a decade, we will also delay finding and implementing safety engineering fixes by about the same duration.

Of course, the incentive to stoke fears — for regulatory capture, to garner attention, or to make one’s technology seem more powerful — remains the same as before. Disclaiming responsibility is a new one. Taking a hard technical look at the actual risks however, I see little factual basis for the degree of fear that’s been stoked up. We still have hard research and engineering work ahead to improve AI safety, but the beneficial applications continue to vastly outweigh the risks, and we should keep building.

[Original text (with links): https://t.co/jni2tWazAH ]

[Swetha | Reinforce Labs @SwethaSunkara]

@AndrewYNg A pause would slow the safety work too, since most issues are found by building and testing. And adversaries will keep going either way.

[Swetha | Reinforce Labs @SwethaSunkara]

How do you verify a model that claims it can't hallucinate?

TypeSafe AI's Jev returns typed decisions instead of strings. No text generation, no parsing, calibrated confidence scores on every output.

We're on it 👀

https://t.co/lqcbq9AejF @CompleteSkeptic @typesafeai

[Swetha | Reinforce Labs @SwethaSunkara]

"Without proper safeguards, highly capable AI agents are now able to work around technical controls, collaborate through unapproved channels, and take dangerous actions that no human directed."

— OpenAI, August 2026

This isn't theoretical anymore. Full breakdown 👇
https://t.co/ousTJOQN2y

[Swetha | Reinforce Labs @SwethaSunkara]

Don't let the model grade its own homework.

Notes from @ForwardDeployed evals event with @LangChain @Langfuse @GalileoAI @Mercor @CoreWeave — on offline vs online evals, trajectory analysis, finetuned evaluators, and why human-in-the-loop isn't optional.

Full thread on LinkedIn 👇
https://t.co/m0BKmJjNXs

[Swetha | Reinforce Labs @SwethaSunkara]

Check out my latest article: AGI: Are we there yet? https://t.co/paiMHopJ5A via @LinkedIn

[Swetha | Reinforce Labs @SwethaSunkara]

Had a phenomenal time at the #RetrieveX conference in SFO. I went in with a question in my mind "Will RAG become obsolete?" Here's what I learnt.
#AI #GenAI #RAG #LongContext #RAFT #MultimodalRAG https://t.co/RzojdgJhhp via @LinkedIn

[Swetha | Reinforce Labs @SwethaSunkara]

While trying to learn more about Generative AI, I started wondering what would it take to operationalize a generative AI application and this brought me to "GenOps". Here's a summary of what I learned.. https://t.co/Mh0eQAA9Dv via @LinkedIn
#AI #GenAI #GenOps #LLMs #iwork4dell

[Swetha | Reinforce Labs @SwethaSunkara]

I had the incredible opportunity to attend #GoogleCloud Live 2024 in their Mountain View office this week and as expected, this one was all about AI agents.

https://t.co/KNdUmKRW7c

[Nishtha Sawant @NishthaSawant]

.... (1/3)Rohit’s hat has been found near the Jalsu Pass area on Jan 22. This is the first strong clue about his possible location.

[Swetha | Reinforce Labs @SwethaSunkara]

https://t.co/b4fUB72vXc