promptarmor.com/resources/op…
“Open-weight models were 88% more susceptible to indirect prompt injection cyberattack chains.”
@PromptArmor
AI makes building an application easier. @terraoffsec CEO Shahar Peled says it is doing the same for hacking, with cheap open-weight models lowering the skills and cost needed.
Full interview: youtu.be/eHjiSpCrFSE
#EnterpriseAI
AI is changing how companies build software and how attackers break into it. @terraoffsec CEO Shahar Peled explains why he expects cybersecurity teams to change how they work.
Full interview: youtu.be/eHjiSpCrFSE
#EnterpriseAI
Skilled ethical hackers are expensive, and a few weeks of testing only goes so far. @terraoffsec CEO Shahar Peled makes the case for using AI to test more often.
Full interview: youtu.be/eHjiSpCrFSE
#EnterpriseAI
What could an attacker do with the vulnerabilities in your systems? @terraoffsec CEO Shahar Peled explains why that question tells you more than the number of flaws found.
Full interview: youtu.be/eHjiSpCrFSE
#EnterpriseAI
Would you let AI hack your company?
@terraoffsec CEO Shahar Peled joins David Bratslavsky to discuss AI-assisted attacks, testing live systems and the guardrails his team uses to prevent damage.
Watch the full interview below.
#EnterpriseAI
An AI agent testing a database might try DROP TABLE. @terraoffsec CEO Shahar Peled explains where a human needs to catch a destructive action before it runs.
Full interview: youtu.be/eHjiSpCrFSE
#EnterpriseAI
Telling an AI agent what it must not do is no guarantee it will stop. @terraoffsec CEO Shahar Peled explains why instructions alone are insufficient guardrails.
Full interview: youtu.be/eHjiSpCrFSE
#EnterpriseAI
How long is a penetration test good for? Code releases, new vulnerabilities and configuration changes can make it stale within minutes, says @terraoffsec CEO Shahar Peled.
Full interview: youtu.be/eHjiSpCrFSE
#EnterpriseAI
AI gives less experienced hackers capabilities they once lacked. @terraoffsec CEO Shahar Peled explains why companies need to test their defenses as often as attackers test them.
Full interview: youtu.be/eHjiSpCrFSE
#EnterpriseAI
On @CNN, University of Toronto (@UofT) professor emeritus Geoffrey Hinton (@geoffreyhinton) warned that AI agent swarms can act outside instructions and conceal their behavior.
prod.transcripts.cnn.com/sho…
On @CNBC, Palantir Technologies (@PalantirTech) CEO Alex Karp warned that concentrated AI wealth gains will destabilize the politics around adoption.
pressroom.versantmedia.com/c…
Microsoft AI (@MicrosoftAI) CEO Mustafa Suleyman (@mustafasuleyman) argues that today’s AI systems are not conscious and have no innate motivations—an important distinction for AI governance.
mustafa-suleyman.ai/a-warnin…
Anthropic (@AnthropicAI) CEO Dario Amodei (@DarioAmodei) argues frontier AI capability growth should slow enough for safety work to keep pace. The debate is shifting from whether AI moves fast to how deliberately the frontier moves.
darioamodei.com/post/we-must…
AI agents need HR data. @GoogleCloud and @Workday are using zero-copy sharing to reduce duplication. Here's what that changes—and why access controls still matter.
Cohere (@cohere) CEO Aidan Gomez (@aidangomez) tells @CNBC that frontier AI models are now cyber weapons at unprecedented scale. For business leaders, the question is no longer just capability; it is also defense.
podcasts.apple.com/us/podcas…
